a constraint-based analysis of passive constructions in korean · 2008-12-28 · 최 재 웅...
TRANSCRIPT
Thesis for the Degree of Master
A Constraint-based Analysis of Passive Constructions in Korean
Sanghoun Song
Majoring in Computational Linguistics
Department of Linguistics
Korea University
December 2007
© 2007 Sanghoun Song
최 재 웅 교수지도
석 사 학 위 논 문
A Constraint-based Analysis of Passive Constructions in Korean
이 논문을 문학석사 학위논문으로 제출함
2007년 12월
고려대학교 대학원
언어학과 전산언어학 전공
송 상 헌
송상헌의 문학석사 학위논문
심사를 완료함.
2006년 12월
위원장 최 재 웅 (인)
위 원 강 범 모 (인)
위 원 김 종 복 (인)
i
Acknowledgements
I have received help from a number of persons during the writing of this
thesis. Moreover, various experience for the past seven years also comes in
pretty hand in completing my thesis. I would like to offer my heartfelt thanks
to those who have afforded me a chance to improve my mind about linguistics.
Prof. Jae-Woong Choe, who always encourages me to study better,
inspired an important thought into me so as to write this thesis. He is my
paragon as a linguist. Prof. Seok-Hoon You is the first person who recognizes
my talent in linguistics. If it had not been for his concern, I couldn't have
devoted myself to linguistics. Prof. Beom-mo Kang made me take an interest
in corpus linguistics. I also received instruction in theoretical linguistics from
Prof. Myung-Yoon Kang and Prof. Sung-Do KIM.
Prof. Suk-Jin Chang and Prof. Kiyong Lee gave me great encouragement
to study HPSG. Prof. Jong-Bok Kim has my eyes opened to HPSG. Thanks
to their help, I could understand the framework of HPSG much better.
For the last two years as a postgraduate student, I have been sponsored by
the second Brain Korea 21 Project – KLING Forum. Dr. Eunjeong Oh has
spared no effort to help me improve my research. Dr. Dong-Sung Kim stirred
up my interest in stochastic processing for language resources.
I had studied at City University of Hong Kong for two months as a
visiting research student. A large part of this thesis was completed during the
visiting period. Prof. Jonathan WEBSTER, the director of Dept. of Chinese,
Translation and Linguistics, showed careful consideration for my visiting. Dr.
Sunyoung Oh helped me adapt myself to the life at City University and gave
advice about my thesis. Ms. Julia HAN allowed me to visit her Korean class,
ii
which made me understand Korean education in foreign countries. Dr. Haihua
PAN, my supervisor during the visiting period, offered me full comments on the
section 6.1.1 of this thesis. I had discussed the computational processing of
language resources with Dr. Alex FANG. His idea furnished me many hints
about natural language processing. I sought counsel from Prof. Benjamin
T'SOU about the prospect of computational linguistics. Prof. Zhiwei Feng
told me a global trend of computational linguistics several times. I took a
lesson in machine translation from Dr. Chunyu KIT. The lesson was a good
chance for me to draw an outline of the foundation of machine translation.
I had worked for Eoneo Inc. before I became a postgraduate student.
My experience in the company has also been helpful to study computational
linguistics. I am thankful to Tosang Chung, their CEO. By his favor, I could
learn how to work expertly and how to make a good plan. I was taught even
Finnish by him. Yeolwon Seong, the senior researcher of natural language
processing team, let me know how to implement natural language processing
systems. Young-ah Kim, the senior researcher, always stood by me at that
time. I’m sure that they will be good friends of mine for aye. I could discuss
linguistic phenomena with Prof. Jung-soo Mok, who had worked for Eoneo Inc.
as a researcher, as well.
Finally, I would give special thanks to my colleagues. Jieun Jeon
always gives me a hand for my study. EUN HO KIM, who is now studying at
University of Hawaii, had helped me with various affairs. Besides, I have had
great and small aid from not a few colleagues.
I am indebted to all of them for what I am. Thanks a lot!
iii
A part of this thesis was presented in the ‘14th International
Conference on HPSG’ held at Stanford University with Prof.
Jae-Woong Choe and published online by CSLI publications
(Song and Choe 2007). I would like to appreciate the
comments from anonymous readers and some members of
the audience during the conference. Thanks to the
comments, I could elaborate this thesis. Of course, all
remaining errors are my responsibility.
iv
Contents
1. Introduction .........................................................................1
1.1 The Goal ..............................................................................................2
1.2 Outline .................................................................................................2
2. Theoretical Background ......................................................4
2.1 HPSG (Head-driven Phrase Structure Grammar) ..................................4
2.1.1 Lexical Category within the HPSG Framework ...........................6
2.1.2 Syntactic Category within the HPSG Framework ........................7
2.1.3 Semantic Category within the HPSG Framework ........................9
2.1.4 The LKB (Linguistic Knowledge Building) System ...................10
2.2 Korean Phrase Structure Grammar ..................................................... 11
2.3 Passives in HPSG ...............................................................................12
2.3.1 Korean Passives in HPSG ..........................................................13
3. Korean Passives .................................................................15
3.1 The Passive Forms .............................................................................15
3.2 The Scope of Korean Passives ............................................................16
3.3 The Taxonomy of Korean Passives .....................................................18
v
3.4 The Function of Passive Markers .......................................................21
3.5 The Data Collection ...........................................................................24
4. Suffixal, Auxiliary, and Inherent Passives ........................25
4.1 Basic Data ..........................................................................................25
4.1.1 Suffixal vs. Auxiliary Passives ...................................................25
4.1.2 Inherent Passives .......................................................................27
4.2 Typed Feature Structures of Passivization ..........................................30
4.3 Passive Types of Verbs .......................................................................32
4.3.1 Classification of Verbs ...............................................................33
4.3.2 Suffixal Passives .......................................................................35
4.3.3 Auxiliary Passives .....................................................................41
4.3.4 Inherent Passives .......................................................................51
5. Verbal Nouns and Passive Light Verbs .............................53
5.1 Basic Data ..........................................................................................53
5.2 Typed Feature Structures of Passivization: Revisited ..........................57
5.3 Passive Types of Verbal Nouns ...........................................................58
5.3.1 Classification of Verbal Nouns ..................................................59
5.3.2 Types of Verbal Nouns and Passive Light Verbs .........................60
vi
5.3.3 Two Kinds of Passive Light Verb Constructions ........................63
6. The Related Constructions ................................................70
6.1 Subtypes of Suffixal Passives .............................................................70
6.1.1 Korean Middles with Passive Suffixes .......................................74
6.1.2 A Spontaneous Event .................................................................81
6.2 Passive-like Forms .............................................................................83
6.2.1 Passive Forms of Intransitive Verbs ...........................................83
6.2.2 Obligatory Passives ...................................................................86
6.3 Double Passives .................................................................................88
7. Conclusion ..........................................................................92
References ..............................................................................94
Appendix I (Classification) ....................................................99
Appendix II (Implementation) ............................................109
vii
Abbreviations
ABL : Ablative case marker
ACC : Accusative case marker
AGT : Agentivity
ARG-ST : Argument-structure
AUX : Auxiliary
AUXPASS : Passive auxiliary / Auxiliary passives
AVM : Attribute Value Matrix
CAUS : Causative suffix
COMP : Complementizer suffix
COMPS : Complements
DAT : Dative case marker
DC : Declarative sentence-type marker
DET : Determiner
GCASE : Grammatical case marker
HFP : Head Feature Principle
HON : Honorific marker
HPSG : Head-driven Phrase Structure Grammar
LEX : Lexical category
LFG : Lexical Functional Grammar
LKB : Linguistic Knowledge Building
LV : Light verb
MM : Middle marker
MOD : Modifier
MORPH : Morphological information
MRS : Minimal Recursion Semantics
NEG : Negative marker
NOM : Nominative case marker
viii
PASS : Passive affix
PASS-TYPE : Passive-type
PAST : Past tense marker
PHON : Phonological information
PLV : Passive light verb
pLVCs : Passive light verb constructions
PLVPASS : Passives with passive light verbs
pLVs : Passive light verbs
POS : Part-of-speech
POTEL : 'Potential' marker
PRES : Present tense marker
REFL : Reflexives
RELN : Relation
RESTR : Restriction
SCASE : Semantic case marker
SEM : Semantics
SPON : 'Spontaneous event' marker
SPR : Specifier
SUBJ : Subject
SUFCAUS : Suffixal causatives
SUFPASS : Suffixal passives
SYN : Syntax
TFS : Typed feature structure
TNS : Tense marker
TOPIC : Topic marker
VALP : Valence Principle
VFORM : Verbal form
VN : Verbal noun
ix
국국국국 문문문문 초초초초 록록록록
본고에서 관심을 두는 언어 현상은 한국어의 피동 구문이다. 피동
구문은 문법 연구 초기에서부터 폭넓게 논의되었으며, 생성문법 등장 이후
에는 문법 기술 및 설명에서 빠질 수 없는 부분으로 인식되어 왔다. 본고에
서 택하는 기본 접근법은 한국어 피동 구문을 유형화하고, 그에 대한 제약
을 밝히며 이 제약을 바탕으로 위계 구조를 도출하는 것이다. 한국어의 기
본 피동구문의 유형에는 크게 접미사 피동, 조동사 피동, 내현적 피동, 피동
형 경동사 구문이 포함된다.
동사의 경우 동사를 피동형의 가능 여부에 따라 다섯 가지로 구분
한다. 다음 단계로 접미사 피동을 용언 위계 구조 속에서 처리하기 위해, 김
종복·양재형 (2006)의 용언 위계구조를 일부 수정하였다. 접미사 피동은 v-
alt-stem 제약에 의해 처리된다. 조동사 피동은 위해 통사 위계 구조에 hd-
lex-pass-ex 제약을 명시하였으며, 관련된 규칙 등도 설정하였다. 한편, '때리
다-맞다'와 같은 어휘쌍이 '능동-피동'의 대립관계를 보임에 주목하여, 내현적
피동사라는 범주를 자동사의 일부로 상정하였다.
서술성 명사의 경우, 한국어에는 '되다', '받다', '당하다' 3개의 피동형
경동사가 존재하는 것으로 보고, 이 경동사와의 공기 제약을 중심으로 유형
x
화하였다. 서술성 명사는 '주어의 유정성 자질', '문법격의 실현', '서술성 명사
의 피해성 자질' 등을 중심으로 하여 총 8개의 유형으로 구분된다. 아울러,
피동형 경동사 각각의 위계 구조와 제약, 그리고 결합을 위한 문법 규칙 등
도 제시된다.
한국어에는 위와 같은 정규적 피동 구문 이외에도 관련된 다양한
현상이 존재한다. 예컨대, 자동사가 피동형을 취하는 경우 등을 들 수 있다.
한국어 피동 구문에 대한 종합적 모델 제시를 지향하는 본고는 이들 범주에
대해서도 제약기반적 분석을 시도한다.
한편으로, 각 유형에 대한 분석 및 검증은 실제 언어 자료에 근간한
귀납적 방법론을 추구하였다는 데 의의를 지닌다.
끝으로, 본고에서 제시된 모든 모델은 유형화된 자질 구조를 사용하
는 문법을 위한 통합 패키지인 LKB에서 모두 구현되어 그 실효성을 검증
받았다.
- 1 -
1. Introduction
The grammatical theory that this thesis is concerned about in order to analyze
language phenomena is the so-called Head-driven Phrase Structure Grammar
(henceforth HPSG). HPSG, one of the surface grammars, has demonstrated
the superiority of its computational process. HPSG generally offers an
explanation of language phenomena on the basis of constraints; therefore the
grammar is often called a constraint-based grammar, which is differentiated
from the so-called derivation-based grammar. Within the HPSG framework,
descriptive exactitude, mathematical logicality, and computational applicability
hold a key position in building up a grammar.1 These characteristics pave the
way to ensure high efficiency in practical application, which is the reason why
HPSG has been employed as a major theoretical framework for natural
language processing for more than a decade.
The phenomena that this thesis covers are the passive constructions in
Korean. The construction of passives is one of the topics that have been
researched widely since the early days of grammar studies, and explanations of
passives may have been regarded as crucial for the validity of a grammatical
theory. Since passives tend to have close correlations with various linguistic
categories, passives occupy a high position in describing a grammar. This
thesis provides an overall explanation of language phenomena in Korean
passives within the HPSG framework, and provides methodology to build up a
grammar from the constraint-based approach.
1 For more information about this, see Pollard and Sag (1994).
- 2 -
1.1 The Goal
The main purpose of this thesis is to provide fine-grained type hierarchies for
passives in Korean within the constraint-based framework. In order to build
up the hierarchies, this thesis makes more articulated analyses of passive
constructions in Korean, and proposes constraints on passives in Korean on the
basis of the analyses. Since computational processing is one of the goals of
HPSG, it is essential to implement the fully implemented grammar into a
practical system to check for computational feasibility. I have implemented
the whole grammar in this thesis in the Linguistic Knowledge Building
(hereafter LKB) system. This thesis ultimately aims to contribute to the
development of a Korean computational grammar, such as the Korean Resource
Grammar (hereafter KRG).
1.2 Outline
Chapter 2 provides an explanation of the theoretical background of this thesis.
To begin with, the HPSG framework will be summarized, and then a brief
history of the HPSG for Korean will be provided. Treatment of passives in
HPSG will also be discussed in chapter 2.
In Chapter 3, the basic assumption on which this thesis stands will be
presented. Section 3.1 defines a criterion for the discrimination of passive
types, section 3.2 defines the scope of Korean passives, section 3.3 systematizes
the taxonomy of Korean passives, and section 3.4 suggests the main function of
passive markers.
Chapter 4 discusses the passivization of verbs. Subcategorizing the
passive forms of verbs into three groups, type hierarchy and constraints for each
of them will be presented. Turning to verbal nouns, chapter 5 will show type
- 3 -
hierarchy of these and of passive light verbs.
Whereas chapter 4 and chapter 5 deal with canonical passives in Korean,
chapter 6 makes an analysis of the non-canonical constructions of Korean
passives. Section 6.1 classifies suffixal passives into three subtypes. The
passive-like forms, such as passive forms of intransitive verbs, will be treated in
section 6.2, and the so-called ‘double passives’ will be discussed in section 6.3.
Chapter 7 is the conclusion of this thesis.
- 4 -
2. Theoretical Background
2.1 HPSG (Head-driven Phrase Structure Grammar)
HPSG is a typical surface grammar for a computational application. Müller
and Sag (2007), which is an introductory lecture note on HPSG, describes the
overview of HPSG as follows.
(1) General Overview of the Framework a. Lexicalized (head-driven) b. Sign-based (Saussure 1916) c. Typed feature structure d. Lexical entries (lexical entries, phrases, principles) e. Multiple inheritance f. Phonology, syntax, and semantics are represented in one description.
First, HPSG belongs to a lexicon-oriented grammar, like Lexical Functional
Grammar (LFG; see Bresnan 1982). HPSG and LFG are often called
lexicalist grammars, because ‘their frameworks move much more information
away from the grammar rules into the lexicon’ (Copestake 2002:116). That is,
HPSG deals with syntactic or semantic categories on the basis of the linguistic
information of lexical categories; therefore lexical units play a significant role
in HPSG. In particular, HPSG lays emphasis on the role of the head-lexicon,
as the name represents. On the other hand, HPSG belongs to the class of
unification-based grammars2, because linguistic information is constructed with
the operation called unification.
Secondly, the model to stands for a lexeme or a word within the HPSG
2 Pollard and Sag (1994:19) regard unification as structure sharing which means ‘two path share the same structure as their common value.’ Furthermore, they suppose it to be a core of HPSG framework.
“It is not going too far to say that in HPSG structure sharing is the central explanatory mechanism, much as move-α is the central explanatory mechanism in GB theory.”
- 5 -
framework is adapted from the conception of Ferdinand de Saussure, who
regarded language as a system of ‘signs’. Although lexical entries in HPSG
are much richer than in Saussure’s conception, the sign is the basic unit for the
representation of linguistic information in HPSG (Sag and Wasow 1999:356).
Furthermore, any sort of linguistic construction, such as words, phrases, or
sentences, in HPSG are regarded as the set of signs, which is very similar to
Saussure’s (1916) idea. (2) from Sag (2007)3 is the typed declaration which
stands for ‘sign’.
(2)
sem-obj
syn-obj
)-structurelist(morph
structure)list(phon-
sign
SEMANTICS
SYNTAX
FORM
PHON
:
One important consideration in contemporary linguistics is to introduce
feature structure into the representation of linguistic information. HPSG
describes the language structure as typed feature structure (hereafter TFS).
TFS is different from feature structure within other theoretical frameworks; TFS
makes use of a more complex and more hierarchical feature system. Roughly
speaking, TFS is a method to express constraints. According to Copestake
(2002:16), “constraints are used to capture generalizations: the type hierarchy
allows for inheritance of constraints.” In other words, each constraint will be
built up hierarchically for linguistic generalizations. What is noticeable about
this hierarchical structure is that a node can inherit information from multiple
upper nodes; multiple inheritance.
Last but not least, one of the distinctive properties of HPSG is that each
3 Sag (2007) proposes namely the Sign-based Construction Grammar which is developed from HPSG.
- 6 -
level of language, such as phonology, morphology, syntax, semantics, or even
pragmatics, is represented synthetically from start to finish.
2.1.1 Lexical Category within the HPSG Framework
A matter of primary concern relating to lexical categories within the HPSG
framework is to reduce redundancy in description. It is possible to build up
lexical categories by listing entries in the lexicon one by one, but there are two
disadvantages in this method. First, this method is uneconomical, because it
leaves overlapping information between lexical items. For example, although
both an intransitive verb and a transitive verb belong to the verb category, the
traditional way to register entries in the dictionary one by one requires
specification of the information that the verb belongs to the verbal category
every time. This kind of redundancy is called ‘vertical redundancy.’
Secondly, this method does not take relationships between words into account.
For example, ‘sing’, ‘sings’, ‘sang’, ‘sung’, ‘singing’, or ‘singer’ are derived
words from the lemma ‘sing’, but the traditional method represents this
paradigm poorly. This is referred to as ‘horizontal redundancy.’
Lexical category in HPSG aims to reduce these redundancies. In order
to cut down on vertical redundancy, type hierarchical system has been
employed in the HPSG framework. On the other hand, lexical rules, such as
the inflectional lexical rule or the derivational lexical rule, have been used to
reduce horizontal redundancy.
Lexical structure in Korean, as described in Kim (2004:36), is shown
below. Basically, this thesis follows the structure below. Verbs treated in
chapter 4 belong to verbal in (3), therefore verbs have a [VERBAL +] feature.
Verbal nouns in chapter 5 multiply inherit lexical information from both verbal
- 7 -
and nominal; verbal nouns have [VERBAL +, NOMINAL +] features.
(3)
2.1.2 Syntactic Category within the HPSG Framework
Syntactic category in HPSG is mainly concerned with constraints on how
expressions can combine with each other. There are two universal principles
in HPSG; the ‘Head Feature Principle’ and the ‘Valence Principle.’ Pollard
and Sag (1994:34) define the former as the following.
(4) The Head Feature Principle (hereafter HFP) The HEAD value of any headed phrase is the structure-shared with the HEAD value of the head daughter.
Schematizing the above definition on Attribute Value Matrix (henceforth
AVM), it will be as the following (Müller and Sag 2007).
(5)
⇒1
1 HEAD|DTRHEAD
HEAD
-aseheaded-phr
The latter, which has been developed from the so-called ‘Subcategorization
Principle’ in the early days of HPSG, can be defined as the following.
(6) The Valence Principle (Sag and Wasow 1999:86, hereafter VALP) Unless the rule says otherwise, the mother’s SPR and COMPS values are identical to those of the head daughter.
- 8 -
(7)
[ ]1 COMPS
[ ]1 COMPS
(8) [ ]1 SPR
[ ]1 SPR
In addition to these principles, there are well-formed phrase structure rules in
HPSG, such as the ‘Head-Subj Rule’, the ‘Head-Comp Rule’, the ‘Head-Mod
Rule’, or the ‘Head-Lex Rule’. The AVMs which stand for each of them are
presented below, which are given in the appendix of Kim (2004).
(9)
Head-Subj Rule
[ ] [ ]11 SUBJ , GCASE|CASE Hnomhd-subj-ph→
(10) Head-Comp Rule A
[ ]
+→... , ..., COMPS
AGT|HEAD , GCASE|CASE
11 Hacchd-comp-ph
Head-Comp Rule B
[ ]
−→... , ..., COMPS
AGT|HEAD , GCASE|CASE
11 Hnomhd-comp-ph
(11) Head-Mod Rule A
[ ]+
→ AGT|HEAD , GCASE|CASE
MOD H1
1acc
-phmodhd-
Head-Mod Rule B
[ ]−
→ AGT|HEAD , GCASE|CASE
MOD H1
1nom
-phmodhd-
- 9 -
(12) Head-Lex Rule
+
+→
1
1 COMPS
AUX ,
L COMPS
LEX
L COMPSH
hd-lex-ex
(13) is the syntactic structure given in Kim (2004:76).
(13)
2.1.3 Semantic Category within the HPSG Framework
HPSG adopts ‘Minimal Recursion Semantics (henceforth MRS)’ as its semantic
representation. Copestake et al. (2005:281) define MRS as (14).
(14) MRS is a framework for computational semantics that is suitable for parsing and generation and that can be implemented in typed feature structure formalisms.
The typed feature structure SEM, which contains semantic information,
generally has the following schema. MODE indicates the semantic type of the
sentences; for example prop, ques, or dir.4 INDEX is a kind of label on
individuals or situations. RESTR, the abbreviation of ‘restriction,’ is to
specify ‘a list of conditions that the situation or individual has to satisfy for the
expression to be applicable to it (Sag and Wasow 1999:106).’
(15)
[ ]
... ...,
index
mode
,RESTR
INDEX
MODE
4 prop is short for ‘proposition, ques is short for ‘question’, and dir is short for ‘directive.’
- 10 -
There are two semantic principles in HPSG. One is the ‘Semantic
Compositionality Principle,’ the other is the ‘Semantic Inheritance Principle.’
(16) and (17) from Sag et al. (2003) define them respectively.
(16) Semantic Compositionality Principle In any well-formed phrase structure, the mother’s RESTR value is the sum of the RESTR values of the daughter.
(17) Semantic Inheritance Principle In any headed phrase, the mother’s MODE and INDEX values are identical to those of the head daughter.
2.1.4 The LKB (Linguistic Knowledge Building) System
The LKB system is a software package for writing linguistic programs
(Copestake 2002). In other words, it is a grammar and lexicon development
environment for TFS grammars5, such as HPSG or LFG. The linguistic
methodology for writing a grammar on the basis of a computational
grammatical theory such as HPSG and implement it into the applicable system
such as LKB is called ‘grammar engineering,’ and this thesis also belongs to the
class of studies in the style of ‘grammar engineering.’6
5 ‘The Lingo project’ is the international research to describe grammars based on the LKB system within the HPSG framework. ERG (English Resource Grammar) is the typical HPSG-based grammar of English. For Korean, KRG is currently under development. 6 Since this thesis intends to provide the basic model of ‘grammar engineering’ as stated before, I have tested all proposals in this thesis in the LKB system.
- 11 -
2.2 Korean Phrase Structure Grammar
An overview of Korean grammar in HPSG is given in Chang (1995) and Kim
(2004).7 They give full details of the core elements of HPSG with regard to
Korean, and further offer an overall description of linguistic phenomena in
Korean from a constraint-based perspective.
It is my understanding that Chang (1995) is the first book to seek to make
comprehensive HPSG-based analyses of Korean. He provides a basic
grammar of Korean with respect to four aspects of that grammar; part-of-speech,
sentence, extended sentence, and discourse. On the basis of these four aspects,
he provides the syntactic parsing and semantic description of Korean sentences
within the HPSG framework.
Since the middle of 90’s, HPSG has developed gradually; for example,
the ‘Subcategorization Principle’ has been substituted with VALP. Within the
developed HPSG framework, Kim (2004) provides an outline of Korean
grammar. He gives an explanation of the foundation of KPSG (Korean Phrase
Structure Grammar) from lexical, syntactic, and semantic perspectives. The
literature treats the major linguistic phenomena in Korean for the most part,
including passives.
One of the purposes of this thesis is to build upon and complement these
previous studies through more articulated analyses. Therefore, I will maintain
the skeleton presented in Chang (1995) and Kim (2004) as far as possible in this
thesis.
7 Park (2004) also presents the basis of HPSG by and large, though he does not deal with Korean data.
- 12 -
2.3 Passives in HPSG
The passive is one of the most frequently analyzed constructions in the tradition
of modern linguistics. Within the constraint-based framework, the passive
construction has been interpreted as a relationship between two verb forms and
a lexical rule has been used to capture generalization, in general. (Sag and
Wasow 1999:233).
Representative research on the passive construction in HPSG can be
found in Müller (2000) and Müller (2003), where lexical-rule based approaches
have been employed in the analysis of the passive.8 Müller (2000) lays
particular emphasis on the morphological processes, because they make it
possible to ‘capture the generalization about passive in a uniform way.’
Building on his idea, I will employ the lexical-rule based analysis with verbal
morphology in this thesis. In order to treat various passive constructions in
Korean in a uniform way, the typed feature structure MORPH that contains the
information about passivization will be used in this thesis.
Ackerman and Webelhuth (1998) provide a full detail of predicates,
including passives, within the HPSG framework. Their claim is relatively
similar to Müller’s (2000) view. Although their study mainly treats German
predicates, their comprehensive approach to verbal expression9 offers a
suggestion for this thesis.
Murphy (2004) proposes type hierarchy for passive constructions.
Murphy’s studies have the goal of investigating passives from a cross-linguistic
standpoint, and present a model of universal passive types within the HPSG 8 According to Müller, the lexical-rule based approach is more suitable for explaining the German passive than the so-called ‘object-to-subject-raising’ analysis. 9 The approach of capturing generalizations about passive constructions on the basis of the whole system of predicate structure is basically similar to those of this thesis. Furthermore, the approach has the same way of analyzing passives with Keenan’s (1985) typological view, in which passives are considered as a way of forming verbal categories.
- 13 -
framework. In line with his approach, this thesis builds up type hierarchies for
passives on the basis of their forms.
2.3.1 Korean Passives in HPSG
As mentioned above, Korean passive constructions have also been a hot topic
since the early days of Korean generative grammar. However, the constraint-
based perspective on Korean passive constructions was introduced only recently,
and there is little literature on the Korean passive in HPSG. Previous research
on Korean passives within HPSG includes Chang (1995), Kim (2005), and Lee
(2005).
Chang (1995) might be the first to have provided an analysis of Korean
passive within the constraint-based framework. In recent years, Kim (2005)
recasts the Korean passive within the HPSG analysis and gives a solution to
computational implementation for it. Lee (2005) shows the significance of
interfaces between syntax and semantics in Korean passives, and suggests a
lexical-rule based approach can be applicable for syntactic passives10. Lee’s
analysis of auxiliary passives or passive light verb constructions is different
from Kim (2005). While Kim (2005) assumes main verbs have a [PASSIVE
+] feature in the auxiliary passive construction, Lee (2005) claims that the
element which is responsible for a [PASSIVE +] feature is not a main verb but a
passive auxiliary.11 This difference of opinion between Kim (2005) and Lee
(2005) will be discussed in section 4.3.3 a little more.
10 According to the terms of this thesis, it is identical to auxiliary passives. The previous studies consistently argue that a lexical-rule based approach is appropriate for not only suffixal passives in Korean but also auxiliary passives in Korean, which is similar to Müller (2000)’s idea. 11 However, Lee (2005) has not come to a conclusion about which one is more meaningful in terms of linguistic or computational usefulness.
- 14 -
These previous studies offer an overall picture of Korean passive
constructions, but they dealt with passives on a rather illustrative basis, showing
that some samples can be handled in HPSG. The purpose of this paper is to
propose more fine-grained type hierarchies for the Korean passive constructions
within the constraint-based grammar.
- 15 -
3. Korean Passives
3.1 The Passive Forms
Keenan (1985:246) argues that linguists who want to study passives should take
a serious view of ‘ways of forming verb phrases,’ because passives belong to
the processes of verbal formations. That is, passive constructions have a close
relation with the formation of a verbal lexicon. In line with his claim, I regard
the forms, particularly the forms of VPs, as an important criterion for Korean
passive types.
Generally speaking, studies within HPSG begin by establishing
classification of types. TFS for constructing linguistic expressions from a
lemma to a sentence stands on the basis of this classification. Therefore, it is
essential for the HPSG-based study to classify linguistic expressions into
subtypes. The main issue in this classification is how to set a criterion. One
can regard syntactic differences as the criterion, or one can take a semantic or
morphological view. Among them, the reason why this thesis places emphasis
on the forms is that classification based on forms can draw the most coherent
TFS with respect to passive constructions.
Haspelmath (1990:27) argues ‘passive constructions without passive
morphology do not exist.’ Yeon (2005:587), likewise, says that morphological
aspects have been disregarded in comparison with syntactic or semantic views
in the study of Korean passives. In other words, any kind of research that does
not consider passive forms will not offer a perfect explanation. In this thesis, I
observe the verbal formation of passive constructions in Korean, and try to
determine the constraints, based on their forms.
- 16 -
3.2 The Scope of Korean Passives
Since there seems to be no clear consensus as to the scope of the passive
constructions in Korean, I give my own definition of Korean passives from a
cross-linguistic perspective.
Keenan (1985:247) regards sentences like (18a) as ‘basic passives.’
(18) a. John was slapped. b. John was slapped by Mary.
(18a) is distinct from other passive or passive-like constructions in three ways.
First, there is no agent phrase that is realized overtly, such as ‘by Mary’ in
(18b). Secondly, the main verb, such as ‘slap’ in (18), is originally transitive.
Finally, the main verb contains an agent subject and a patient or theme object in
its active sentence. (19) is Keenan’s generalization on basic passives.
(19) G-2: If a language has any passives it has ones characterized as basic passives; moreover, it may have only basic passives. G-2.1: If a language has passives with agent phrases then it has them without agent phrases. G-2.2: If a language has passives of stative verbs then it has passives of activity verbs. G-2.3: If a language has passives of intransitive verbs then it has passives of transitive verbs.
Let me make a similar application of this typological generalization to passive
constructions in Korean. To begin with, the appearance of agent phrases12 in
regular passive sentences is optional. Since adjectives in Korean generally 12 Haspelmath (1990:56) says ‘passives that are lexically restricted or passives that are rarely used seldom allow the expression of an agent, whereas very general and very frequent passives, as in English and in some Bantu languages, freely allow the expression of the agent.’ Korean passives are midway between them; though the appearance of agent phrases is not limited; upon investigation into Korean corpora, such as the Sejong Korean Treebank, passive sentences without agent phrases are more common than those with agent phrases.
- 17 -
play the part of stative verbs, G-2.2 in (19) need not be considered.13 Above
all, this thesis assumes that the canonical passive type in Korean falls under the
conjugation of transitive verbs, because it is highly exceptional that intransitive
verbs can be derived into passives in contemporary Korean.
In addition to this, this thesis regards morphological properties as one of
the major criteria for distinguishing whether a type of construction is under the
scope of passives or not. As mentioned in the previous section, Haspelmath
(1990:62) claims ‘the passive can usefully be regarded as being primarily a
morphological verbal category.’ That is, passivization basically refers to the
transformation of active forms into passive forms.
From these points of view, this thesis adopts the following assumption as
the scope of canonical passives in Korean. Hereafter, all analyses to Korean
passives are grounded on (20).
(20) The scope of Korean passives a. In principle, only transitive verbs can be transformed into passives. The
passive sentence, therefore, must have both agent and theme roles, though the agent role may not be realized overtly.
b. There should be a corresponding active form for each passive form. Besides, passives must be morphologically distinct from their corresponding actives.
Of course, (20) is not uncontroversial, because there are some
counterexamples even in contemporary Korean. Contra (20a), there are
several intransitive verbs that can be transformed into passives or passive-like
forms; for example, nal- ‘fly’ or col- ‘drowse’.14 In the opposite case of (20b),
13 As a matter of course, there are Korean verbs that contain stative property; for example, talm- ‘resemble’. They share general properties with stative verbs in other languages; they are seldom transformed into progressives, imperatives, or passives. The issue of these verbs will be discussed in chapter 4. 14 Kwon (1993) argues that intransitive verbs could be passivized more freely in Middle
- 18 -
there are some passive forms for which there is no corresponding active. For
example, cong-i wul-li-ta ‘the bell rings’ has no corresponding active form. In
this example, wul-li- morphologically belongs to the passives, but the verb
functions like an active. This characteristic is more or less similar to the so-
called ‘deponent verbs’ in Latin (Kiparsky 2005).
In spite of these counterexamples, I would regard them as non-canonical
passives or passive-like constructions. This is on account of two reasons.
One is that constructions like those rarely appear in comparison with canonical
passives of (20), and the verbs used for the constructions are highly restricted.
The other stands in line with Kemmer’s (1993a) claim. She argues that
deponent verbs, such as (cong-i) wul-li-ta ‘(the bell) rings’, are cross-
linguistically universal, and if there is no example of deponent verbs in any
kind of investigation into passives in a language, it is because the investigation
is not perfect yet. Building upon her claim, it is never strange that there are
some cases that go against the scope of canonical passives. Hence, this thesis
deals with canonical passives in conformity with (20) first, and then tries to
make an analysis of non-canonical passives in chapter 6.
3.3 The Taxonomy of Korean Passives
Passive constructions in Korean are traditionally divided into three subgroups;
suffixal passives, auxiliary passives, and passive light verb constructions.15
Korean in the 15th century. As mentioned in G-2.3 of (19), this phenomenon actually is not so peculiar. Keenan (1985:250) cites an instance in Latin; currere ‘to run’, curritur ‘it is run’. 15 Chae (2003) proposes similar classification.
“There are two different types of passive constructions in Korean: lexical passives, those containing a verb with the passive morpheme -i, -hi, -li or -ki, and syntactic passives. Syntactic passives are, in turn, divided into two groups: those formed by the auxiliary verb ci-, which takes an -e form complement, and those containing a Verbal Noun (VN) and a passive Light Verb (pLV) like toy-, pat- or tangha-.”
- 19 -
Suffixal passives are expressed by suffixes whose occurrence is conditioned
largely by the stem-final sounds. There are four variants in the suffix; -i, -hi, -
li , and -ki. For example, ccic- ‘tear’ takes the suffix -ki to form a passive verb
like ccic-ki- ‘be torn’. Auxiliary passives are phrasal passives which consist of
a verbal stem followed by the complementizer16 -e or -a and the auxiliary -ci as
in ccic-e ci- ‘be torn’. Passive Light Verb Constructions (henceforth pLVCs,
named after Chae 2003) consist of verbal nouns (hereafter VNs) and passive
light verbs (hereafter pLVs), such as toy-, pat-, and tangha-. For instance, the
active light verb construction, such as chepel ha- ‘punish’ which is made up of
a verbal noun chepel ‘punishment’ and a light verb ha-, can be transformed into
the passives, as in chepel toy-/pat-/tangha- ‘be punished.’17 Sample sentences
related to each of them are given below in sequence.
(21) ku os-i ccic-ki-ess-ta. DET dress-NOM tear-PASS-PAST-DC ‘The dress was torn.’
(22) ku os-i ccic-e ci-ess-ta. DET dress-NOM tear-COMP AUXPASS-PAST-DC ‘The dress was torn.’
(23) Mia-ka chepel-toy-ess-ta. Mia-ka chepel-pat-ass-ta. Mia-ka chepel-tangha-yess-ta. Mia-NOM punishment-PLV-PAST-DC. ‘Mia was punished.’
In addition to the above subgroups of Korean passives, this thesis includes
16 This term may be more or less controversial because the auxiliary passive construction is not a complex clause in a general sense. I tentatively define the suffix that attaches to main verbs as COMP in this thesis, which is similar to a COMP of Chang (1995:16) or a complementizer affix of Kim (2004:52). 17 There are some opposite views to this taxonomy (see Sung 1976, Im 1998).
- 20 -
verbs which are called lexical passive verbs by Cho (1998) and Sohn (1999)18
as one of the passive constructions. For example, mac- ‘be hit’ shows a
contrast with ttayli- ‘hit’, therefore these two verbs stand on the relationship of
passive/active correspondence.
(24) a. Inho-ka Mia-lul ttayli-ess-ta. Inho-NOM Mia-ACC hit-PAST-DC ‘Inho hit Mia.’ b. Mia-ka Inho-eykey mac-ass-ta. Mia-NOM Inho-DAT be hit-PAST-DC ‘Mia was hit by Inho.’
According to Cho (1998), mac- in (24b) may be analyzed as a passive in a
broad sense, in terms of its passive-like meanings and syntactic behavior. This
thesis includes verbs like mac- among Korean passives, because the verb that
belongs to this type conforms to (20), which was presented as a criterion for
Korean passives. First, mac- and ttayli- in (24) behave like a passive/active
pair in terms of their argument structure. The agent role in (24a) corresponds
to the theme role in (24b), the theme role in (24b) corresponds to the agent role
in (24b). Secondly, mac- and ttayli- are distinctive from each other formally,
though mac- does not contain any markers. Considering that verbs like mac-
have passive information from the start, I will call this type ‘inherent passives’
from now on. The more concrete reason why this type should be considered
as one of the subtypes of Korean passives will be discussed in section 4.1.2.
In a nutshell, I classify Korean passive expressions into four subclasses,
including the inherent case. The taxonomy of canonical passives in Korean is
sketched out below. 18 Sohn (1999:371) divides lexical passives into two subgroups. One is identical with pLVCs according to the view taken in this thesis. The other is equivalent to an inherent passive verb by definition of this thesis. He classifies Korean passives into three groups; suffixal, lexical, and phrasal passives. The last ones are just the same as auxiliary passives.
- 21 -
(25)
3.4 The Function of Passive Markers
Since passive expressions generally contain passive markers, the forms play an
important role in the characterization of passives, as aforesaid. Therefore, the
emphasis of this thesis is on passive markers making a passive formation. The
question that I would like to raise here is about what the main function of
passive markers is.
Before answering the question, let me refer to previous studies to
consider the relationship between passive suffixes in Korean and causative
suffixes in Korean. As is well known, Korean suffixal passives and suffixal
causatives share the suffixes; -i, -hi, -li , and -ki.19 Therefore, (26) from Yeon
(2002:198) can be interpreted ambiguously.
(26) John-i ipalsa-eykey meli-lul kkakk-i-ess-ta. John-NOM barber-DAT hair-ACC cut-CAUS/PASS-PAST-DC a. ‘John made the barber cut his hair.’ b. ‘John had/got his hair cut by the barber.’
In this context, there have been several researches on the relationship
between passives and causatives in Korean. Yang (1979) argues that suffixal
passives and suffixal causatives are of the same type. His claim may look
19 Three more suffixes are used for causatives in Korean other than these; -wu, -kwu, and -chwu. They are usually used for the causative forms of adjectives, but they are not always so. For instance, the causative form of kkay-ta ‘wake up’ is kkay-wu-ta.
- 22 -
somewhat unintelligible, but morphological similarity between them affords a
solid basis for it. In fact, if we investigate the verbs that can be derived into
both suffixal passives and suffixal causatives, we will find out that the same
suffix is used for the verb’s transformation, almost invariably.20 Yang (1979)
provides the whole list to prove this similarity, and furthermore supports his
claim from a phonological, syntactic, and semantic perspective. More recently,
Kang (1997) pays attention to the relationship between suffixal passives and
suffixal causatives, and gives a solution to it in terms of bare phrase structure
within the minimalism of Chomsky (1995). He introduces the so-called vp-
shell into his suggestion. Sohn (1999:367) takes notice of the same
morphological slot of the passive suffix and causative suffix. According his
proposal, ‘Korean passive suffixes have developed from causative suffixes via
functional shift.’
Though these previous studies21 are based on different theoretical
backgrounds, it is obvious that there is a relationship between suffixal passives
and suffixal causatives. They have something in common; both suffixal
passives and suffixal causatives belong to a marked form. That is, there is a
markedness which distinguishes suffixal passives or suffixal causatives from
their corresponding neutral verb.
Turning to the issue raised before, what is the main function of passive
markers? If an expression is constructed markedly, it generally implies that
the expression performs a special function as can be supposed from (27).
20 Only one exception presented on Yang (1979)’s list is mek- ‘eat’; mek-hi- ‘be eaten’, mek-i- ‘let someone eat’. 21 In addition to these previous studies, Dr. Sunyoung Oh at City University of Hong Kong (personal communication) said that one of the reasons why suffixal passives and suffixal causatives use the sound [i], such as -i,-hi, -li , and -ki, is that [i] sound has a high distinctiveness in the Korean vowel inventory. In other words, suffixal passives or suffixal causatives phonologically belong to marked form in Korean, which is analogous to my viewpoint that passives are marked forms of verbal expression in order to indicate argument alternation.
- 23 -
(27) “Categories that are cognitively marked tend also to be structurally marked.” (Givón 1991:106)
This thesis stands on the assumption that the major function of passive
markers, such as suffixes -i, -hi, -li , and -ki, an auxiliary ci-, or pLVs toy-, pat-,
and tangha-, is to lead argument alternation. Sohn (1999:368) presents the
procedure of passivization in Korean as (28).22
(28)
a. The object (accusative) of the sentence is made the subject (nominative).
b. The active verb is replaced by the matching passive verb. c. The subject (nominative) of the active sentence is made the agent
with a locative/goal particle or otherwise omitted. d. The new subject is placed at the sentence-initial position.
Syntactically or semantically, the argument structure of an active sentence can
be altered by passive markers which represent the subject of the sentence being
the target of the event structure, rather than an actor. Passive markers, that is,
take the role of a trigger to switch the corresponding pair between the
subject/object and the agent/theme role. Speaking from a functional
perspective, passive markers indicate that the sentence has been derived
throughout argument alternation.
Kim (2004:199), in this context, regards passive constructions as one of
argument alternation structures along with causatives. This thesis is also
based on the assumption that argument alternation is the kernel of passive
structures.
22 (28) may be a universal procedure in world languages, except for (28c).
- 24 -
3.5 The Data Collection
I have attempted to consider the range of relevant data for our studies in a
systematic and comprehensive way, because I believe that the data-oriented
approach works for describing the characteristics of language much better. In
order to collect relevant data, I took advantage of four linguistic resources: the
Sejong POS-tagged Corpora23 , the Sejong Electronic Dictionary24 , the
Standard Korean Dictionary25, and the Yonsei Korean Dictionary26. In
sections 4.3.1 and 5.3.1, I will give a full detail of the process of data
compilation for this study.
In order to extract, sort out, and classify data automatically, I have
implemented several computer programs coded in the Python27 programming
language.
23 These morpheme-tagged corpora include approximately ten-million “words,” or graphic words which are called eojeol in Korean. 24 The version used for this study contains 18,618 verbal surface forms. 25 There are approximately five hundred thousand entries in this dictionary. 26 The verbal category in this dictionary covers 49,552 entries altogether. 27 Python, an interpreter language, may be more convenient for this kind of linguistic-oriented research than other languages which need compilation, such as C/C++, C#, or Java. Furthermore, simplicity in text-processing within Python modules provides strong support for handling large corpora.
- 25 -
4. Suffixal, Auxiliary, and Inherent Passives
4.1 Basic Data
4.1.1 Suffixal vs. Auxiliary Passives
Typical passive forms of verb in Korean contain suffixes like -i, -hi, -li and -ki,
and therefore the active/passive correspondence has been treated either as part
of a syntactic process or as a lexical redundancy rule. However there are a
large number of exceptions to this generalization, and this should be taken into
account. For example, there is no suffixal passive counterpart of the active
verb mandul- ‘make’, as shown in (29).
(29) a. Mia-ka ku sangca-lul mandul-ess-ta. Mia-NOM DET box-ACC make-PAST-DC ‘Mia made the box.’ b. *ku sangca-ka mandul-li -ess-ta. DET box-NOM make-PASS-PAST-DC ‘The box was made.’ c. ku sangca-ka mandul-e ci-ess-ta. DET box-NOM make-COMP AUXPASS-PAST-DC ‘The box was made.’
There is no expression like (29b), because some verbs like mandul- cannot be
used as passives with suffix; *mandul-li- is ill-formed. Whereas verbs like
mandul- cannot combine with a passive suffix, verbs like tat- ‘close’28 are the
opposite with reference to passivization. Though the auxiliary passive
construction is a more productive operation than the suffixal passive
construction, certain verbs sound odd when passivized in this way.29
28 In Korean, tat- ‘close’ is used only as a transitive verb, unlike English. 29 It could be somewhat controversial to make a comparison between the less productive one and the more productive one. But the primary goal of this paper is to draw an outline of the
- 26 -
(30) a. Mia-ka ku sangca-lul tat-ass-ta. Mia-NOM DET box-ACC close-PAST-DC ‘Mia closed the box.’ b. ku sangca-ka tat-hi-ess-ta. DET box-NOM close-PASS-PAST-DC ‘The box was closed.’ c. ??ku sangca-ka tat-a ci-ess-ta. DET box-NOM close-COMP AUXPASS-PAST-DC ‘The box was closed.’
(30c) sounds awkward, while the suffixal passive predicate in (30b) which
corresponds to (30a) sounds perfect. However, as given below, some verbs
like ccic- ‘tear’ can be passivized with an auxiliary verb as well as with a suffix.
(31) a. Mia-ka ku os-ul ccic-ess-ta. Mia-NOM DET dress-ACC tear-PAST-DC ‘Mia tore the dress.’ b. ku os-i ccic-ki-ess-ta. DET dress-NOM tear-PASS-PAST-DC ‘The dress was torn.’ c. ku os-i ccic-e ci-ess-ta. DET dress-NOM tear-COMP AUXPASS-PAST-DC ‘The dress was torn.’
The most serious problem on this restriction is that it is actually
unpredictable whether a verb can be passivized with suffixes or an auxiliary.
Lee (2004:89) discovers that even the set of verbs that convey a similar
meaning do not share distributional properties of passive types, presenting
evidence given in table (1). Asterisks in table (1) mean there is no
corresponding suffixal passive form, such as *mandul-li- in (29b).
Korean passive system within the constraint-based framework. Therefore, it is necessary to discuss the difference in form concerning passivization, which is one of the main properties of Korean passives.
- 27 -
Table (1) ACTIVE PASSIVE ACTIVE PASSIVE ccic- ‘tear’ ccic-ki- pwuswu- ‘smash’ * tam- ‘put a thing into’ tam-ki- neh- ‘put a thing into’ *
pokk- ‘roast’ Pokk-i- salm- ‘boil’ * an- ‘hug’ an-ki- phwum- ‘hug’ *
In order to solve this problem, there is no option but to specify
information about passivization onto the lexicon. It is possible to specify the
possibility of passivization for each verb, but a more efficient way to encode the
same information would be to make use of type hierarchy. It would also
ensure a more natural and systematic grouping of verbs.
4.1.2 Inherent Passives
Inherent passive verbs30, as discussed previously, mean cases in which a verb
contains a [PASSIVE +] feature inherently. Inherent passive verbs do not
include any passive morpheme on the surface, but do show passive-active
correspondence semantically. For the sake of convenience, (24) in section 3.3
is rewritten in (32) below.
(32) a. Inho-ka Mia-lul ttayli-ess-ta. Inho-NOM Mia-ACC hit -PAST-DC ‘Inho hit Mia.’ b. Mia-ka Inho-eykey mac-ass-ta. Mia-NOM Inho-DAT be hit-PAST-DC ‘Mia was hit by Inho.’
30 This type may be controversial because the lexicons that belong to this type are highly restricted. Besides, it is possible to regard the so-called inherent passive verbs as one of the intransitive verbs that require dative cases. Nevertheless, this thesis would assume that there are inherent cases in Korean passive paradigm, taking notice of the possibility of passivization.
- 28 -
It is noticeable that verbs in (32) can be passivized neither with the passive
suffixes nor with passive auxiliary, such as *ttayli-i-, *ttayli-e ci-, *mac-hi-,
and *mac-a ci-.
If it is allowable that the relationship between mac- and ttayli- is a
passive/active pair, mac- in (32b) may be the suppletive form31 of ttayli-, just
as, in English, the comparative form of ‘good’ is not ‘*gooder’ but ‘better’, or
the past tense of ‘go’ is not ‘*goed’ but ‘went’. From a cross-linguistic
viewpoint, it has been said that suppletive expressions of morphosyntactic
forms show a tendency to block the use of morphologically regular expressions.
Application of this typological fact to ttayli- and mac- will result in an
interesting argument. Since ttayli- in (32a) has a suppletive passive form, it is
incompatible with both passive suffixes and a passive auxiliary. mac- in (32b),
likewise, is not compatible with any kind of passivization, just as ‘*betterer’ or
‘*wented’ are ill-formed forms in English. ‘better’ in English has information
about the comparative already and ‘went’ has a [+PAST] feature inherently.
The lexeme mac- similarly has linguistic information about passivization
inherently. Hence, it is reasonable to introduce ‘inherent passives’ into the
taxonomy of passive constructions in Korean.
I consider the ill-formed forms, such as *ttayli-i-, *ttayli-e ci-, *mac-hi-,
and *mac-a ci-, as a consequence of the ‘blocking effect’. Poser (1992:111)
says ‘blocking is the widely observed phenomenon where the existence of one
form prevents the creation and use of another form that would otherwise be
expected to occur.’32 Kiparsky (2005:113) also claims that ‘blocking is a
filtering device that applies to the output of the generative system.’ Kiparsky
31 Fromkin et al. (2003:596) define ‘suppletive forms’ as ‘inflected morphemes in which the regular do not apply.’ 32 According to Poser’s discussion, ‘the non-suppletive forms were phrasal and only suppletive forms are lexical.’ Referring to Sohn (1999) who defines verbs like mac- as lexical passives, the above demonstration may be more convincing.
- 29 -
provides a lexicalist analysis of inflectional paradigms in Latin passives. What
is interesting with reference to inherent passives in Korean among his data is
the so-called ‘activa tantum.’ The verbs called ‘activa tantum’ cannot take a
passive morpheme in the present tense, and other active verbs serve as the
missing passive verbs, as given in (34).
(33) Regular passive inflection in Latin a. amare- ‘to love’ b. amatur ‘is loved’ c. laudat ‘praises’ d. laudātur ‘is praised’
(34) Activa tantum in Latin a. perdi- ‘to destroy’ b. periit ‘perishes, is destroyed’ c. facit ‘makes’ d. fit ‘becomes, is made’
Kiparsky claims that activa tantum like facit in (34c) are inherently specified as
[PASSIVE −], therefore they cannot combine with a passive ending; while
amatur in (33b) is well-formed, *facitur is not. The relationship between facit
and fit in (34c-d) is similar to that of ttayli- and mac- in (32), except in relation
to tense.
If so, it is necessary to think of a device to handle blocking. The device
will be able to treat differences in the manner of passivization presented in
section 4.1.1 as well; some verbs allow only suffixal passives, some verbs allow
only auxiliary passives, some verbs allow both of them, and some verbs
disallow both. The following section gives a basic solution to block incorrect
passive forms within the HPSG framework, in particular, using TFS.
- 30 -
4.2 Typed Feature Structures of Passivization
From basic data in the previous section, I came to two conclusions. One is
that TFS to make it possible to discriminate between well-formed forms and ill-
formed forms is needed. The TFS has to play the function of blocking
passivization with suffixes for mandul- ‘make’ in (29), and of blocking any
kind of passivization for mac- ‘be hit’ in (32b), for instance. In this thesis, I
propose PASS-TYPE as a subitem of MORPH that contains linguistic
information about morphological variation.
(35)
...
...
...
... TYPECAUS
AUXPASS
SUFPASS TYPEPASS
MORPH -
boolean
boolean-
The other is related to a [PASSIVE ±] feature. Transitive verbs have a
[PASSIVE −] feature before argument alternation, and then the feature value
becomes positive in the procedure of argument alternation. Intransitive verbs
normally have a [PASSIVE −] feature, too. Granting that there are some
exceptional cases as stated in section 3.2, this thesis is based on the assumption
that intransitive verbs with a [PASSIVE −] feature cannot participate in passives
in conformity with (20); therefore, the feature value will not be changed.
However, mac- is inherently specified as [PASSIVE +], in contrast with
ordinary verbs.
- 31 -
(36)
_EE
_ER
RELN
RESTR
INDEX
SEM
]NP[,om]NP[ STARG
PASSIVE|HEAD|SYN
−
j
i
s
accn-
v-np-tr
ji
X
X
X
(37)
]NP[ STARG
AUXPASS
SUFPASS TYPEPASS|MORPH
PASSIVE
−−
−
nom-
-
intrv-basic-
(38)
_EE
_ER
RELN
RESTR
INDEX
SEM
]NP[,]NP[ STARG
AUXPASS
SUFPASS TYPEPASS|MORPH
PASSIVE
−−
+
i
j
s
accnom-
-
assive-vinherent-p
ji
X
X
X
- 32 -
4.3 Passive Types of Verbs
I suggest that verbs in Korean are initially classified into four subtypes with
respect to passivization, excluding the ones that do not allow any kind of
passivization like talm- ‘resemble’. The subtypes are primarily differentiated
from each other according to whether they allow only one of the suffixal and
auxiliary types of passivization, or both.
(39)
Verbs which allow only suffixal passives belong to (p-1) type. (p-2) type
involves the verbs that can be transformed into passives only by auxiliary verbs.
Verbs of (p-3) type allow both types of passives. Then there is the other
possibility where a verb allows neither suffixal nor auxiliary passives. We
will call it type (p-4). Examples of the four types are given in the following
table, where the bold faced verbs indicate the blocked forms.
Table (2) SUFPASS AUXPASS (p-1) tat-ta ‘close’ tat-hi-ta ??tat-a ci-ta (p-2) mandul-ta ‘make’ *mandul-li-ta mandul-e ci-ta
(p-3) ccic-ta ‘tear’ ccic-ki-ta ccic-e ci-ta (p-4) talm-ta ‘resemble’ * talm-ki-ta ??talm-a ci-ta
The last line in the table should be distinguished from the cases of the
inherent passive verbs like mac- ‘be hit’. It allows neither passivization
process, but it still has its passive counterpart, albeit inherent, so I suggest that
the inherent passive forms its own type. The overall picture of verb types is
- 33 -
sketched out in (40).
(40)
v-intr... ... inherent-pass-v
v-lxmv-np-tr
v-tr-p1 v-tr-p2 v-tr-p3 v-tr-p4
tat-talm-ttayli- mac-
SUFPASS +AUXPASS -
mandul-
SUFPASS -AUXPASS +
ccic-
SUFPASS +AUXPASS +
SUFPASS -AUXPASS -
PASSIVE +SUFPASS -AUXPASS -
The feature specification in (40) shows clearly which kind of passivization is
allowed for each type. It also shows whether a given type is inherently
passive.
4.3.1 Classification of Verbs
I started from the Sejong POS-tagged Corpora to get the list of verbs which
have a frequency over nine. There were 2,557 verbs on the initial list. Let us
call it ‘List A.’ Next I extracted the suffixal passive forms from each of the
Sejong Electronic Dictionary, the Standard Korean Dictionary, and the Yonsei
Korean Dictionary. Avoiding marginal or controversial cases, I included only
the forms which are admitted to be suffixal passives in all of the dictionaries.
Finally, I excluded from our suffixal passives list the items whose
corresponding active forms are not on List A. As a result, there were 167
suffixal passive forms collected in this way.
As for auxiliary passives, I searched the Sejong POS-tagged Corpora to
- 34 -
find phrasal forms such as ‘V-e/a ci-.’ There were 785 types of verbs which
appeared in this context. From this list, I excluded some cases through the
following four processes. First, the list of verbs were checked against the
Standard Korean Dictionary to determine which one have the suffixal passive
forms (e.g. po-i-ta, ‘be seen’) or causatives (e.g. pes-ki-ta, ‘take off other’s
clothes’), as listed in the dictionary. In this way, non-active forms were
excluded from the 785 types. Second, those verbs which have adjective usage
(e.g. palk-ta, ‘be bright’) were also discarded, because an adjective combined
with ‘-e/a ci-’ has an inchoative33 meaning as in palk-a ci-ta ‘brighten’. Third,
I got rid of intransitive verbs from the data. Fourth, I also got rid of the verbs
which have a locative case-mark alternation, such as NP[loc]-ey/lul hyangha-ta
‘go towards NP[loc].’ Finally, I excluded the items which are not on List A.
Consequently, I obtained 402 verbs that can be passivized with an auxiliary.
In accordance with taxonomy mentioned in section 3.3, I rearranged verbs
entries and classified them into three subcategories. Some examples are given
in (41)34, and the whole list will be presented in the Appendix.
(41)
(p-1) tat-ta ‘close’, mek-ta ‘eat’, ssu-ta ‘use’, cap-ta ‘catch’ (98 verbs) (p-2) mandul-ta ‘make’, kus-ta ‘draw’, is-ta ‘link’, chac-ta ‘find’ (333 verbs) (p-3) ccic-ta ‘tear’, ssu-ta ‘write’, ssis-ta ‘wash’, phwul-ta ‘solve’ (69 verbs)
33 In this thesis, an inchoative predicate refers to a predicate that describes a change of state. 34 In (41), there is a homonym; ssu- ‘use’, ssu- ‘write’. The former belongs to type (p-1), while the latter belongs to type (p-3). That means it is impossible to predict the way a verb will passivize phonologically. Since no generalization is possible, it is conceivable to mark the lexicon with information about passivization. As mentioned in 4.1.1, specifying information for each entry is an uneconomical method as well as tedious work within the HPSG framework, because the motivation for lexical category in HPSG is to reduce redundancy. Hence, this thesis makes use of the type hierarchy and lexical-rule based approach, which are used for decreasing redundancy.
- 35 -
4.3.2 Suffixal Passives
Cho and Sells (1995) and Sells (1995) propose a Korean morphological system
within the HPSG framework. (42) shows the general order of the Korean
verbal paradigm, and (43) from Sells (1995: 314) presents Korean verbal
morphology. That is, type hierarchy is used in order to capture generalizations
about the combination of verbal suffixes.
(42) Vroot – Honorific (V1) – Tense (V2) – Mood (V3) – Discourse (V4) (43)
=TYPE~
read'' PRED
V
[ ]TYPE~
V
+=TYPE~
HON
V1
[ ]TYPE~
V
=TYPE~
FUT TNS
V2
=ROOT : TYPE
DECL MOOD
V3
==
+==
ROOT: TYPE
DECL MOOD
FUT TNS
HON
read'' PRED
V
The above approach has since developed, and recently Kim and Yang (2006)
proposed a type hierarchy (45) for sequences like (44a) which shows the typical
structure of a verb with its suffixes including passive or causative suffixes.
- 36 -
However, notice that passives and causatives are not properly represented in the
hierarchy.35
(44) a. V-base + (PASS/CAUS) + (HON) + (TNS) + MOOD + (COMP) b. cap-hi-si-ess-ta ‘catch-PASS-HON-PAST-DC’
(45)
As is well known, a derivational suffix is inclined to combine with the base
before an inflectional suffix does. That is, inflection takes place after
derivation. A passive suffix, therefore, can be followed by other suffixes in
morphological system as shown in (44). In this context, the basic approach of
Cho and Sells (1995:134) regarding suffixal passives or causatives is that ‘the
derivational morphology, such as passive or causative suffixes, generates new
roots from old roots.’ That means cap-hi- ‘be caught’ in (44a) is passed up to
the inflectional level.
In contrast, the approach of this thesis is that it is unnecessary to classify
suffixes as derivational or inflectional with respect to passive suffixes.
Korean passive or causative suffixes basically show the characteristics of
derivational suffixes. But they function like an inflectional suffix, because
they lead argument alternations. Because of this morphosyntactic peculiarity
of the Korean verbal system, the question of whether passive suffixes are
derivational or inflectional has been a hot issue for a long time, reflecting how 35 The sequence MOOD + (COMP) in (44a) is treated as forming a v-free node in (45).
- 37 -
difficult it is to draw a strict line between derivational suffixes and inflectional
suffixes.
The previous studies that provide support for my argument are as follows:
Kim (1992) classifies Korean verbal suffix into three subgroups; inflection,
derivation, and inflectional derivation. In line with his claim, passive suffixes
belong to the last one. Sohn (1999) also said that passive or causative suffixes
in Korean are on the border between inflection and derivation. Lee (1999:193)
assumes that there exists a hemi-category between inflection and derivation in
Korean verbal system, and suggests the concept titled the ‘stem formation
suffix’, without making a conclusive remark.
Since I also believe Korean passive or causative suffixes are midway
between derivation and inflection, I handle passive suffixes in the type
hierarchy of verbal morphology. (46) is the revised verbal hierarchy which is
able to treat a suffixal passive verb properly within the verbal system.
Reflecting the previous discussion that the main function of passive markers is
argument alternation, I name the super-class of suffixal passives v-alt-stem.
The node v-alt-stem is inserted between v-hon-stem and v-lxm in the type
hierarchy.
(46)
verbalv-stem vn
v-freev-tns-stemv-tnsv-hon-stem
v-honv-lxm
v-alt-stemv-alt
v-pass-i, -hi, -li, -ki
v-caus
- 38 -
(47) presents a lexical rule that shows the actual derivation of passive
forms. If the stem has the features [PASSIVE –] and [PASS-TYPE.SUFPASS
+], it can turn into a v-pass type with the appropriate suffix. The process of
argument alternation will take place between the values of the two ARG-ST
features in (47) as shown in the crossed linking relations of the arguments,
represented as i and j.
(47)
+−
+⇒
1
1
SEM
)](NP[ ,NP STARG
SEM
NP ,NP STARG
SSTYPE.SUFPAPASS
PASSIVE STEM
PASSIVE
idat
j-
ji-
-
v-tr
v-pass
(48) is an AVM which stands for tat- ‘close’, and (49) shows how (p-1)
type like tat-hi- ‘be closed’ is derived. As shown in (48), PASS-TYPE of type
v-tr-p1 is specified as [SUFPASS +, AUXPASS −]. Therefore, passivization
with the suffix -hi will be successful, while passivization with a passive
auxiliary ci- will be blocked.
- 39 -
(48)
−+
−
j
i
s
anom-
-
tat
ji
ARG2
ARG1
close RELN
RESTR
INDEX
SEM
cc]NP[ ,]NP[ STARG
AUXPASS
SUFPASSTYPEPASS
PASSIVE
(49)
+→
−+
−
2
1
2
1
SEM
)](NP[ ,NP STARG
PASSIVE
PHON
ARG2
ARG1
INDEX
SEM
NP,NP STARG
AUXPASS
SUFPASS TYPEPASS
PASSIVE
PHON1
idat
j-
tat-hi
v-pass
j
i
sji
-
-
tat-v-tr-p
For the sake of clarity, the whole tree diagram of (50) is sketched out
below.
- 40 -
(50) ku sangca-ka tat-hi-ess-ta. DET box-NOM close-PASS-PAST-DC ‘The box was closed.’
(51)
NP_SUB VP
v-freev-tns
-ta-ess
-hi
tat
[ ]nom CASE.GCASE2
−
321
SEM
, STARG
PASSIVE
1
-
v-tr-p
+
312
SEM
][, STARG
PASSIVE
dat-
v-pass
argument alternation
+
12
COMPS
SUBJ VAL
PASSIVE
+
1 COMPS
SUBJ VAL
PASSIVE
φ
VP
head-subj-rule
ku sangca-ka
- 41 -
4.3.3 Auxiliary Passives
The hierarchy of syntactic structure below is from Kim (2004:76).
(52)
hd-lex-ex includes most complex predicates36 ; for instance, auxiliary
constructions, light verb constructions, etc. The most significant point of this
constraint is that two verbal expressions combine with each other at the stage of
lexical category, not at the syntactic stage. That is, while categories under
phrase in (52), such as hd-subj-ph, are constructed syntactically, hd-lex-ex is
not. In hd-lex-ex, the combination of two verbal expressions will result in a
new single verbal expression; the new expression has not become a phrase yet.
The rewrite rule (53) and the AVM (54) from Kim (2004:123) show the
property of hd-lex-ex.
(53) V´ → V1 V2
(54)
+
+→
1
1 COMPS
AUX ,
L COMPS
LEX
L COMPSH
hd-lex-ex
It is possible to capture two generalizations from AVM (54). One is that, in
the ‘V + AUX’ frame, AUX is the head of the construction and it takes V as its
complement. The other is that the mother-category inherits COMPS from the
main verb directly, which is marked as ‘L’ on the AVM. 36 For more information about them, see Alsina et al. (1996) for cross-linguistic research or Kim and Yang (2004) for Korean. Both of them are grounded on a constraint-based perspective.
- 42 -
Kim (2005) proposes that auxiliary passives can be handled as hd-lex-ex.
What is notable here are the conditions under which argument alternation takes
place. He suggests that main verbs or verbal nouns cannot combine with an
auxiliary such as ci- or passive light verbs such as toy-, pat-, and tangha- until
they are transformed into passives. This is because the feature structure
COMPS of the main verb should be transmitted to the V´ without any
alternation, as shown in (54). That means argument alternation should be
completed before combining with a passive auxiliary or pLVs. (55) is adapted
from Kim (2005:83).
(55)
31 COMPS
SUBJ
+
32
COMPS
PASSIVE
VFORM HEAD
ae
VNP, ST-ARG
COMPS
SUBJ
V
2121
I agree that auxiliary passives should fall under hd-lex-ex. However,
this thesis suggests hd-lex-pass-ex as one of subtypes of hd-lex-ex, because it is
necessary for hd-lex-ex to branch out. There are two reasons for this.
First, let us consider which conveys the sense of passives. Is it the main
verb or the passive auxiliary ci-? The main claim of this thesis is that hd-lex-
pass-ex should be introduced as a subtype of hd-lex-ex, because what is
responsible for passive meaning is not the verb but the passive auxiliary ci-.
- 43 -
Hence, the process of argument alternation is triggered only when the verb
combines with ci-, thus making the system more controlled.37 In my opinion,
this approach has the added benefit of getting rid of an extra process of vacuous
case alternation for every verb.
Secondly, there are several other uses of Korean -e/a ci- construction
other than passive constructions. If -e/a ci- phrase combines with adjectives, it
represents an inchoative meaning like (56a). On the other hand, if -e/a ci-
phrase combines with forms already passivized as in (56b), it would seem to
convey some resultative meaning.38 These phenomena raise the necessity to
classify -e/a ci- phrases into several subtypes.
(56) a. Mia-ka yeypp-e ci-ess-ta. Mia-NOM pretty-COMP AUX -PAST-DC ‘Mia became pretty.’ b. ku saken-i ic-hi-e ci-ess-ta.39 DET incident-NOM forget-PASS-COMP AUX -PAST-DC ‘The incident has been forgotten.’
(57) is the revised syntactic hierarchy that I propose for the auxiliary
passive construction.
37 It is similar to Lee (2005), who claims argument alternation does not take place until the main verb combines with the passive auxiliary ci-. 38 Of course, (56b) is different from resultative constructions in Korean in a general sense. Here, the concept of ‘resultative’ refers to the result after undergoing the complete verbal action, which is adapted from Haspelmath (1993b). General properties of Korean resultative constructions are well-discussed in Kim and Maling (1998) or Lee and Lee (2002). 39 In a school grammar, this kind of ‘double passive’ form is considered to be wrong, but this form is used far more frequently than the more ‘correct’ form ice-hi-ess-ta. More discussion will be presented in section 6.3.
- 44 -
(57)
That is, hd-lex-ex is divided into two subgroups. One is for the constructions
that need argument alternation, like passives; the other is for constructions that
are not accompanied by any alternation. Argument inheritance of the latter
hd-lex-nonalt-ex is the same as that of (54); the mother-category inherits
COMPS from the main verb without alternation. hd-lex-pass-ex belongs to the
former hd-lex-alt-ex where argument alternation can occur.
The new constraints that this thesis proposes are as follows.
(58)
[ ] [ ]
++→
11
COMPS
AUX , LEX Hhd-lex-ex
(59)
( )
+
+→
22112
SUBJ
PASSIVE,
, STARG
SSTYPE.AUXPAPASS
COMPS
SUBJ
V
-
-
v-comps-exhd-lex-pas
(58) is the constraint for hd-lex-ex, replacing (54) above, and (59) is the rule for
the auxiliary passive construction. In contrast with (54), (58) ignores the
constraint on inheritance of COMPS. The constraint on concrete inheritance
will be specified in more particular types, according to the implementation of
- 45 -
argument alternation.
(60) is the lexical representation for mandul- ‘make’, (61) is the lexical
representation for the passive auxiliary ci-, and (62) shows how the appropriate
passive forms for type (p-2) are derived. Since type (p-2) has a [SUFPASS –]
feature, the suffixal passivization process will be blocked.
(60)
+−
−
j
i
s
anom-
-
mandul
ji
ARG2
ARG1
make RELN
RESTR
INDEX
SEM
cc]NP[ ,]NP[ STARG
AUXPASS
SUFPASSTYPEPASS
PASSIVE
(61)
−
++
PASSIVE
VFORM ],NP[ STARG
PASSIVE
AUX
aenom-
aux-pass
ci
- 46 -
(62)
( )
+
12
4
COMPS
SUBJ
PASSIVE
HEAD
PHON
'
ci-ess-tamaldul-e
V
+−
−
21
3
, STARGAUXPASS
SUFPASS TYPEPASS
PASSIVE
2 STEM PHON
-
-
v-tr-pmaldul-e
v-comp
++
32
4
COMPS
SUBJ
AUX
PASSIVE
PHON ci-ess-ta
V
(64) is the whole derivation diagram of (63).
- 47 -
(63) ku sangca-ka mandul-e ci-ess-ta. DET box-NOM make-COMP AUXPASS-PAST-DC ‘The box was made.’
(64)
NP_SUB VP
mandul-e
argument alternation
VP
head-subj-rule
+
SEM
COMPS
SUBJ VAL
PASSIVE
41
2
s-exhd-lex-pas
V´
−
SEM
, ST-ARG
PASSIVE
421
3
v-comp
++
COMPS
SUBJ
PASSIVE
AUX
32
V
ci-ess-ta
[ ]nom CASE.GCASE2
ku sanca-ka
+
1 COMPS
SUBJ VAL
PASSIVE
φ
- 48 -
(66) and (67) illustrate how (p-3) type (p-3) verbs like ccic- ‘tear’ in
(65) is are derived. Since both PASS-TYPE features of ccic- are plus, ccic-
can be transformed into either ccic-ki- or ccic-e ci-.
(65)
++
−
ij datnom-
-
v-tr-p
ccic
]NP[ ,]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
PASSIVE
3
(66)
+→
+ PASSIVE
PHON
SSTYPE.SUFPAPASS
PHON
3
ccic-ki
v-pass
-
ccic-
v-tr-p
11
(67)
+ PASSIVE
PHON
'
ci-ess-taccic-e
V
+SSTYPE.AUXPAPASS
3 STEM
PHON
-
v-tr-p
ccic-e
v-comp
++
AUX
PASSIVE
PHON ci-ess-ta
V
Likewise, sample derivations are sketched out in (69) and (71). They
are different from the previous diagrams; both of them are satisfied with
COMPS. The top node of (51) for tat-hi- ‘be closed’ or (64) for mandul-e ci-
‘be made’ is not named ‘S’ but ‘VP’, because the complements of them have
not yet become saturated. If all elements of COMPS as well as SUBJ are
eliminated, the sentence will be parsed completely.
- 49 -
(68) ku os-i Mia-eykey ccic-ki-ess-ta. DET dress-NOM Mia-DAT tear-PASS-PAST-DC ‘The dress was torn.’
(69)
NP_DAT VP
v-freev-tns
-ta-ess
-ki
ccic-
+
312
SEM
][, STARG
PASSIVE
dat-
v-pass
argument alternation
+
12
COMPS
SUBJ VAL
PASSIVE
VP
head-comp-rule
Mia-eykey
−
321
SEM
, STARG
PASSIVE
3
-
v-tr-p
S
NP_SUB[ ]nom CASE.GCASE2
ku os-i
[ ]dat CASE.SCASE1
head-subj-rule[ ]+ PASSIVE
+
φ COMPS
SUBJ VAL
PASSIVE
2
- 50 -
(70) ku os-i Mia-eyeky ccic-e ci-ess-ta. DET dress-NOM Mia-DAT tear-COMP AUXPASS-PAST-DC ‘The dress was torn.’
(71)
NP_DAT VP
ccic-e
argument alternation
VP
head-comp-rule
+
SEM
COMPS
SUBJ VAL
PASSIVE
41
2
s-exhd-lex-pas
V´
−
SEM
, ST-ARG
PASSIVE
421
3
v-comp
++
COMPS
SUBJ
PASSIVE
AUX
32
V
ci-ess-ta
S
NP_SUB[ ]nom CASE.GCASE2
ku os-i
Mia-eykey
[ ]dat CASE.SCASE1
head-subj-rule[ ]+ PASSIVE
+
φ COMPS
SUBJ VAL
PASSIVE
2
- 51 -
4.3.4 Inherent Passives
As previously discussed, inherent passive verbs need to have passive
information from the start. Further information must be specified to block the
passive rules from applying to them. AVM (72) and (73) are the lexical
representations for the inherent passive verb mac- ‘be hit’ and for the
corresponding active verb ttayli- ‘hit’, respectively. SUFPASS and
AUXPASS, specified only with a negative value in (72) and (73), will block
any kind of passivization. Verbs that cannot be passivized like talm-
‘resemble’ also belong to v-tr-p4.
(72)
−−
+
ij datnom-
-
ass-vinherent-p
mac
]NP[ ,]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
PASSIVE
(73)
−−
−
ji accnom-
-
v-tr-p
ttayli
]NP[ ,]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
PASSIVE
4
(74)
−−
−
ji accnom-
-
v-tr-p
talm
]NP[ ,]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
PASSIVE
4
- 52 -
(75) Mia-ka Inho-eykey mac-ass-ta. Mia-NOM Inho-DAT be hit-PAST-DC ‘Mia was hit by Inho.’
(76)
[ ]nom CASE.GCASE1
+21, STARG
PASSIVE
-
assive-vinherent-p
[ ]dat CASE.SCASE2
[ ]+ PASSIVE
+
φ COMPS
SUBJ VAL
PASSIVE
1
+
21
COMPS
SUBJ VAL
PASSIVE
- 53 -
5. Verbal Nouns and Passive Light Verbs
5.1 Basic Data
Chae (2003) shows three passive light verbs in Korean; toy-, pat-, and tangha-.
The point that this chapter is concerned with is that there are co-occurrence
restrictions between VNs and pLVs.40 For example, the pLV toy- can attach
to cheypho ‘arrest’ to form the passive verb cheypho-toy- ‘be arrested,’ but the
same VN cheypho with another pLV pat-, such as *cheypho-pat-ta, is not a
legitimate form in Korean as shown below.
(77) a. kyengchal-i Mia-lul cheypho-ha-yess-ta. policeman-NOM Mia-ACC arrest-LV -PAST-DC ‘A policeman arrested Mia.’ b. Mia-ka kyengchal-eykey cheypho-toy-ess-ta. *Mia-ka kyengchal-eykey cheypho-pat-ass-ta. Mia-ka kyengchal-eykey cheypho-tangha-yess-ta. Mia-NOM policeman-DAT arrest-PLV -PAST-DC ‘Mia was arrested by a policeman.’
The constraint on pLVCs also determines which nominal can be taken as
the complement of a pLV. Korean light verb constructions have a case frame
such as ‘VN(-ul/lul[ACC]) + ha.’ The passive forms for the frame can be
divided into two forms; ‘VN(-i/ka[NOM]) + PLV’ or ‘VN(- ul/lul[ACC]) +
PLV.’ (78) shows the difference between them clearly.
(78) a. kyengchal-i Mia-lul cheypho-lul ha-yess-ta. policeman-NOM Mia-ACC arrest-ACC LV-PAST-DC ‘A policeman arrested Mia.’
40 This thesis skips over the general issues on light verbs or verbal nouns. For more information, see other literature. The major issues on light verb constructions in Korean are discussed by and large in Chae (1996). Kang (2001) provides a detailed explanation to semantic structures of verbal nouns.
- 54 -
b. Mia-ka kyengchal-eykey cheypho-ka toy-ess-ta. *Mia-ka kyengchal-eykey cheypho-ka tangha-yess-ta. Mia-NOM policeman-DAT arrest-NOM PLV -PAST-DC ‘Mia was arrested by a policeman. c. *Mia-ka kyengchal-eykey cheypho-lul toy-ess-ta. Mia-ka kyengchal-eykey cheypho-lul tangha-yess-ta. Mia-NOM policeman-DAT arrest-ACC PLV -PAST-DC ‘Mia was arrested by a policeman.
There are three forms of passivization for VNs, and the pLVs are distinct
from each other with respect to the choice of VNs. The meanings of the three
pLVs are different from each other as well. Basically, toy- means ‘become,’
pat- may convey a sense of ‘reception,’ and tangha- can be translated into
English as ‘suffer.’ The sentences in (79) are cases in which toy-, pat-, and
tangha- are made use of as main verbs with their regular verbal meanings.
(79) a. Mia-ka kyoswu-ka toy-ess-ta. Mia-NOM professor-NOM become-PAST-DC ‘Mia became a professor.’ b. Mia-ka pyenci-lul pat-ass-ta. Mia-NOM letter-ACC receive-PAST-DC ‘Mia received a letter.’ c. Mia-ka sako-lul tangha-yess-ta. Mia-NOM accident-ACC suffer-PAST-DC ‘Mia suffered an accident. (Mia met with an accident.)’
As noted in footnote 17, there are some opinions against my claim that
pLVCs belong to canonical passives in Korean. In spite of their claims, in this
thesis, I will consider pLVCs as one of the Korean passive constructions, from a
typological perspective. Keenan (1985:257) says that there are four types of
periphrastic passives. According to his analysis, periphrastic passives may fall
into natural subclasses depending on the choice of the auxiliary verb: ‘being’ or
- 55 -
‘becoming,’ ‘reception,’ ‘motion,’ or ‘experience.’
(80) a. Hans wurde von seinem Vater besttaft. Hans became ‘by’ his father punished ‘Hans was punished by his father.’ (German, Keenan 1985:257) b. Mia-ka chepel-i toy-ess-ta. Mia-NOM punishment-NOM become-PAST-DC ‘Mia was punished.’
(81) a. Cafodd Wyn ei rybuddio gan Ifor. get Wyn his warnings by Ifor ‘Wyn was warned by Ifor.’ (Welsh, Keenan 1985:259) b. Mia-ka chepel-ul pat-ass-ta. Mia-NOM punishment-NOM receive-PAST-DC ‘Mia was punished.
(82) a. Quang bi (Bao) ghet. Quang suffer (Bao) detest ‘Quang is detested (by Bao).’ (Vietnamese, Keenan 1985:260) b. Mia-ka chepel-ul tangha-yess-ta. Mia-NOM punishment-NOM suffer-PAST-DC ‘Mia was punished.
Haspelmath (1990:38) similarly divides passive auxiliaries into two groups.
One includes intransitive auxiliaries, such as, ‘be’, ‘become’, ‘stay’, ‘come’,
and ‘go’. For example, in Ecuadorian Quechua, riku-ri-n ‘it is seen’ is derived
from the verb ri- ‘go’ which functions like a passive marker. The other is the
set of transitive auxiliaries, such as, ‘undergo’, ‘come’, and ‘receive’.41 (83) is
the well-known bèi construction in Mandarin, which belongs to the latter
according to the Haspelmath’s explanation.
(83) Tā bèi (tàitai) kànjiàn. he PASS wife see. ‘He is seen (by [his] wife).’
41 Haspelmath (1990) says that the former is much more frequent than the latter.
- 56 -
In accordance with Haspelmath’s view, toy- ‘become’ is one of the
intransitive auxiliaries, while pat- ‘receive’ or tangha- ‘suffer’ belong to the set
of transitive auxiliaries. The facts mentioned in the above bear out my
argument that toy-, pat-, or tangha- can be used for passive constructions.
The next question may be why there are three elements. If I must give
an answer to it in advance, it is because toy-, pat-, and tangha- convey a
distinctive sense, respectively. Data in Malagasy, presented in Keenan
(1985:252), offers strong support for this answer. An active sentence (84a)
corresponds to all of (84b-d). That means one active sentence can be
transformed into three passive sentences.
(84) a. Man+tsangana(=manangana) ny lai aho. ACT+put up the tent I. ‘I am putting up the tent.’ b. A-tsanga-ko ny lai. PASS-put up-by me the tent ‘The tent is put up by me.’ c. Voa-tsangana ny lai. PASS-put up the tent ‘The tent is put up.’ d. Tofa-tsangana ny lai. PASS-put up the tent ‘The tent is put up.’
Keenan clarifies the distinction between (84b-d). a-, voa-, and tofa- in (84b-d)
are differentiated from each other semantically; (84b) is the neutral passive,
(84c) conveys a perfective meaning, and (84d) denotes a spontaneous event.
From this cross-linguistic viewpoint, it is not surprising that there are
three elements in Korean pLVs. Moreover, it is certain that the difference in
meaning between them gives rise to co-occurrence restrictions. The main
concern of this chapter is how to capture generalizations with regard to the co-
- 57 -
occurrence restrictions on the relationship between the VN and the pLVs.
5.2 Typed Feature Structures of Passivization: Revisited
There are also some VNs that cannot be members of pLVCs, as exemplified in
(85). (85b) sounds highly odd, and (85c-d) seem ungrammatical.
(85) a. Mia-ka enehak-ul swuhak-ha-yess-ta. Mia-NOM linguistics-ACC study-LV-PAST-DC ‘Mia studied linguistics.’ b. ??enehak-i Mia-eykey swuhak-toy-ess-ta. linguistics-NOM Mia-DAT study-PLV -PAST-DC ‘Linguistics was studied by Mia.’ c. *enehak-i Mia-eykey swuhak-pat-ass-ta. linguistics-NOM Mia-DAT study-PLV -PAST-DC ‘Linguistics was studied by Mia.’ d. *enehak-i Mia-eykey swuhak-tangha-yess-ta. linguistics-NOM Mia-DAT study-PLV -PAST-DC ‘Linguistics was studied by Mia.’
In order to block unacceptable forms such as ??swuhak-toy- in (85b),
AVM (35) should be revised as (86). The PLVPASS feature for discriminating
between well-formed pLVCs and ill-formed pLVCs is added into PASS-TYPE.
(86)
...
PLVPASS
AUXPASS
SUFPASS
TYPEPASS MORPH
boolean
boolean
boolean
-
- 58 -
5.3 Passive Types of Verbal Nouns
VNs also constitute a type hierarchy of their own with respect to their
combination with light verbs. Therefore, this thesis also proposes a type
hierarchy of VNs in relation to pLVs.
Since there are three pLVs available for combination with VNs, there are
eight types of VNs with respect to passivization, including a case such as
swuhak ‘study’ where a VN cannot take any pLVs.
(87)
It is surprising that actual verbal nouns for each of the logically possible seven
types are attested in Korean. Asterisks in Table (3) show the unacceptable
forms.
Table (3) + toy- + pat- + tangha- (vn-1) chepel ‘punishment’ O O O (vn-2) yongse ‘forgiveness’ O O *
(vn-3) cheypho ‘arrestment’ O * O (vn-4) myelsi ‘contempt’ * O O
(vn-5) yenkwu ‘research’ O * * (vn-6) conkyeng ‘respect’ * O *
(vn-7) paysin ‘betrayal’ * * O
(vn-8) swuhak ‘study’ * * *
- 59 -
5.3.1 Classification of Verbal Nouns
The major diagnostic criterion for VNs is whether a given noun can be
combined with the light verb ha- (Kang 2001). Therefore, I first extracted
from the Sejong Electronic Dictionary (2002-3) a list of nouns whose lexical
entries specify that they can be combined with ha-. I further consulted with
the entries of the VN items on the list and narrowed the list to those items
whose entry specifies that they have the case frame of ‘NP-ul/lul VN-ul/lul ha.’
This restriction was introduced to ensure that the nontransitive VNs are
excluded because they cannot have a passive counterpart in principle. I also
excluded the cases where the VN consists of one syllable or over three syllables
because these tend to involve some semantic peculiarity. The resulting
number of VNs was 2,707. The next step was to find positive evidence for
possible combination of VNs and pLVs by searching the Sejong POS-tagged
Corpora. For instance, given a VN yenkwu ‘inquiry’ and a pLV toy-, I
searched the corpus to see whether there is a form similar to yenkwu-toy in the
corpus. Likewise, I also checked for sequences such as VN-pat-, VN-lul pat-,
VN-tangha-, VN-lul tangha- in the corpus. Altogether 1,713 (or 1,595 if more
strict criteria are adopted) VNs out of the 2,707 were found to be combinable
with one or more of the three pLVs.
(88) (vn-1) chepel ‘punishment’, kangyo ‘forcible demand’, ekap ‘suppression’,, chwukwung ‘pressing hard’, chimhay ‘infringement’ (54 VNs / 30 VNs) (vn-2) yongse ‘forgiveness’, taychwul ‘loan’, poko ‘briefing’, sangsok ‘inheritance’, sentayk ‘selection’ (264 VNs / 189 VNs) (vn-3) cheypho ‘arrestment’, kamkum ‘imprisonment’, salhay ‘murder’, apswu ‘confiscation’, hayko ‘dismissal’ (120 VNs / 122VNs) (vn-4) myelsi ‘contempt’, chimlyak ‘invasion’ (29 VNs / 6 VN)
- 60 -
(vn-5) yenkwu ‘inquiry’, naptuk ‘assent’, tunglok ‘registration’, pannap ‘return’, punsil ‘loss’, sayong ‘use’, kangjo ‘emphasis’, yoyak ‘summation’, insang ‘raising’, cunpi ‘preliminary’, hoypok ‘recovery’ (1,030 VNs / 1,127 VNs) (vn-6) conkyeng ‘respect’, daywu ‘respect’, senmang ‘envy’, chingchan ‘praise’, hoanyeng ‘welcome’ (160 VNs / 74 VNs) (vn-7) paysin ‘betrayal’, sahyeng ‘punishment of death’, hoksa ‘abuse’, kwutha ‘blow’, kangkan ‘rape’ (56 VNs / 47 VNs)
The numbers following the slash at the end of each type are the resulting
number of cases where ‘VN+ul/lul pLV’, such as chepel-ul pat- or chepel-ul
tangha-, are excluded when searching the corpus.42 Incidentally, there were
994/1,112 VNs for which there was no case of pLV passivization found in the
corpus, including swuhak ‘study’, kikwuen ‘abstention’, paywung ‘send-off’,
poksup ‘review’, etc.
5.3.2 Types of Verbal Nouns and Passive Light Verbs
In the case of pLVCs, inductive reasoning based on the data observed above led
me to the conclusion that there are three features which seem to be relevant to
their restrictions.
First, the ‘animacy’ of the subject seems to be relevant. The
constructions with pat- or tangha- are inclined to have an animate subject.
Secondly, the grammatical cases of VNs is also relevant. toy- takes a
nominative case noun as its complement, whereas pat- or tangha- take an
accusative.43 Finally, ‘adversity’ feature of VNs seems to play a role.
42 Type (vn-4) has the fewest number of examples; apparently, if a VN can be combined with a pLV, but cannot be combined with -toy, then -pat and -tangha are in complementary distribution. 43 For this purpose I can make use of the feature AGT, introduced by Kim (2004:152). (10) in chapter 2 shows the function of AGT. If a complement is assigned the nominative case, the
- 61 -
Almost invariably, tangha- combines with nouns which convey a sense of
adversity. To take an instance, cheypho ‘arrest’ comes under the category of
event which brings the patient into an adverse situation. On the other hand,
the event that conkyeng ‘respect’ denotes never does injury to the patient.
Therefore, cheypho can combine with tangha- freely, whereas conkyeng cannot.
Table (4) shows the overall picture of these phenomena.
Table (4) toy- pat- tangha- ANIMATE − + + AGT − + + ADVERSITY bool − +
After considering all these factors, I have built up a type hierarchy for VNs.
(89)
complement can only combine with predicates with [AGT −]. On the other hand, predicates with [AGT +] can take only accusative as their complement.
- 62 -
Table (4) and type hierarchy (89) are also compatible with earlier literature44
that has researched the relationship between VNs and pLVs.
cheypho ‘arrest’ has an unspecified ANIMATE feature as well as
[ADVERSITY +], for instance. According to the constraint, cheypho can
combine with toy- or tangha-. The lexical representation for cheypho is given
in (90). The category VN is represented by the features [POS noun,
NOMINAL +, VERBAL +].45
(90)
+
+
++
j
i
arrest
s
--
bool
noun
vn-tr-plv
cheypho
ji
ARG1
ARG1
RELN
RESTR
INDEX
SEM
NP ,NP STARG SSTYPE.PLVPAPASS
ADVERSITY
ANIMATE
VERBAL
NOMINAL
POS
HEAD|SYN
3
(91) shows a type hierarchy for PLVs. The upper dotted box represents
a difference in the ANIMATE feature value and grammatical cases. The lower
one stands for the difference in the ADVERSITY feature value.
44 For more information, see Lee (2001), Kim (2002), Chae(2003), etc. 45 For more information about Korean VNs within the HPSG framework, see Choi and Wechsler (2001), Kim et al. (2004), or Ryu (2006).
- 63 -
(91)
5.3.3 Two Kinds of Passive Light Verb Constructions
In chapter 4, I handled suffixal passives within type hierarchy for verbal
morphology, and made use of a syntactic operation for auxiliary passives. The
methods are applicable to (92), similarly.
(92) a. Mia-ka kyengchal-eykey cheypho-toy-ess-ta. Mia-NOM policeman-DAT arrest-PLV-PAST-DC ‘Mia was arrested by a policeman.’ b. Mia-ka kyengchal-eykey cheypho-lul tangha-yess-ta. Mia-NOM policeman-DAT arrest-ACC PLV-PAST-DC ‘Mia was arrested by a policeman.’
cheypho-toy- in (92a) is constructed morphologically. It is analogous to the
suffixal passive construction ccic-ki- ‘be torn’. cheypho-lul tangha- in (92b),
on the other hand, is parallel with an auxiliary passive construction ccic-e ci-
‘be torn’. That is, (92a) can be parsed within the morphological system,
whereas (92b) can be analyzed as one of subtypes under hd-lex-alt-ex.
First, let me explain how cheypho-toy- ‘be arrested’ can be derived. vn-
- 64 -
pass-light-v in (93) is the super-class of inflectional items -toy, -pat, and -
tangha. According to constraint (93), vn-pass-light-v can only attach to the
type vn-tr with a [PLVPASS +] feature. Thanks to this device, VNs with a
[PLVPASS −] feature, such as swukak ‘study’ in (85), cannot combine with any
kind of pLVs. (94), (95), and (96) stand for -toy, -pat, and -tangha, in
sequence.46 They reflect conditions given in table (4) except for AGT. AGT
need not be considered here, because this construction is not built up at the
phrasal level. After combining with pLVs, the new stem can be followed by
other verbal suffixes, such as v-hon, v-tns, or v-free in (46).
(93)
+
+⇒
1
1
SEM
)](NP[ ,NP STARG
SEM
NP ,NP STARG
SSTYPE.PLVPAPASS STEM
PASSIVE
idat
j-
ji-
-
vn-tr
ght-vvn-pass-li
(94) [ ][ ] ANIMATE STEM −
toy-
(95)
−
+ ADVERSITY
ANIMATE STEM
-pat
46 That means there are two entries for each of them. One is specified on the inventory of inflectional units, the other is written down on the dictionary along with a passive auxiliary ci-.
- 65 -
(96)
+
+ ADVERSITY
ANIMATE STEM
tangha-
(97) Mia-ka kyengchal-eykey cheypho-toy-ess-ta. Mia-NOM policeman-DAT arrest-PLV-PAST-DC ‘Mia was arrested by a policeman.’
(98)
NP_DAT VP
v-freev-tns
-ta-ess
-toy
cheypho
+
312
SEM
][, STARG
PASSIVE
dat-
ght-vvn-pass-li
argument alternation
+
12
COMPS
SUBJ VAL
PASSIVE
VP
head-comp-rule
kyengchal-eykey
S
NP_SUB[ ]nom CASE.GCASE2
Mia-ka
[ ]dat CASE.SCASE1
head-subj-rule[ ]+ PASSIVE
+
φ COMPS
SUBJ VAL
PASSIVE
2
+
321
SEM
, STARG
ANIMATE
PLVPASS
3
-
bool
vn-tr-plv
- 66 -
Secondly, this thesis suggests hd-lex-vn-pass-ex as a subtype of hd-lex-
alt-ex in (57).
(99)
The representations of pLVs will be as follows. Each name was given before
in (91).
(100)
[ ]+⇒
PASSIVE
-vpass-light
(101)
−
−⇒
ANIMATE
AGT
-toypass-light
(102)
+
+⇒
ANIMATE
AGT
-anipass-light
(103)
[ ]−⇒
ADVERSITY
-patpass-light
(104)
[ ]+⇒
ADVERSITY
tangha-pass-light
With constraints and rules in (105), (106), and (107), the second pLVC can be
dealt with properly. These AVMs reflect the key features discussed so far;
- 67 -
animacy, adversity, and grammatical cases. Besides, since hd-lex-vn-pass-ex
in (105) requires that the first daughter should have a [PLVPASS +] feature, the
constraint can perform a function to block illegitimate forms such as ??swuhak-
toy- ‘be studied’ in (85b).
(105)
[ ]
+
++
→ ADVERSITY
ANIMATE,
SSTYPE.PLVPAPASS
ADVERSITY
ANIMATE
VERBAL
NOMINAL
POS
21
21
-
noun
pass-exhd-lex-vn-
(106) head-lex-vn-pass-rule-1
[ ]
−→1
1 COMPS
AGT, GCASE][ nompass-exhd-lex-vn-
(107) head-lex-vn-pass-rule-2
[ ] [ ]
+→
+ 11
COMPS
AGT, GCASE
ANIMATE SUBJacc
pass-exhd-lex-vn-
- 68 -
Sample derivations for the second pLVC are given below.
(108)
+
12 COMPS
SUBJ PASSIVE
PHON'
gha-ass-tatan lcheypho-luV
−+
++
32 , STARG
AGT
ADVERSITY
ANIMATE
PHON gha-ass-tatan
V
−
+
21
3
, STARG
CASE.GCASE
ADVERSITY
ANIMATE
PHON
acc
bool
lcheypho-lu
VNP
cheypho ‘arrest’ has an unspecified ANIMATE feature value, [ADVERSITY +],
and [PLVPASS +]. Therefore, it can combine with pLV tangha- which has an
[ANIMATE +] as well as an [ADVERSITY +]. Even if an accusative case is
allocated to cheypho, pLVCs will be constructed without any problem thanks to
an [AGT +] of tangha-. Finally, the whole picture of the derivation will be as
(110).
- 69 -
(109) Mia-ka kyengchal-eykey cheypho-lul tangha-yess-ta. Mia-NOM policeman-DAT arrest-ACC PLV-PAST-DC ‘Mia was arrested by a policeman.’
(110)
NP_DAT VP
cheypho-lul
argument alternation
VP
head-comp-rule
V´
tangha-yess-ta
S
NP_SUB
Mia-ka
kyengchal-eykey
[ ]dat CASE.SCASE1
head-subj-rule[ ]+ PASSIVE
++
++
+
COMPS
SUBJ
AGT
ADVERSITY
ANIMATE
PASSIVE
AUX
32
V
[ ][ ]
+
+
SEM
COMPS
ANIMATE SUBJ SUBJ VAL
PASSIVE
41
2
pass-exhd-lex-vn-
+nom CASE.GCASE
ANIMATE2
+
φ COMPS
SUBJ VAL
PASSIVE
2
+
−
SEM
, ST-ARG GCASE
ADVERSITY
ANIMATE
PASSIVE
421
3acc
bool
VNP
- 70 -
6. The Related Constructions
One of the purposes of this thesis is to propose synthetic and comprehensive
analyses to Korean passives within the constraint-based grammar. The issues
discussed so far come under the category of canonical passives in Korean. On
the one hand these analyses form the foundation of parsing Korean passive
sentences, on the other hand they are still lacking something. It is clear that
the previous discussion covers most of passive constructions in Korean, but it
has not yet given the solution for some marginal cases; for example, the passive
formation of intransitive verbs. In order to provide a fine-grained grammar for
passives to the extent that it is available for a practical system, it is important to
deal with related constructions. This chapter aims to offer solutions to the so-
called passive-like constructions or rather controversial topics on Korean
passives.
6.1 Subtypes of Suffixal Passives
Let us first look at passive sentences in Malagasy once again. For
convenience, the data is rewritten below.
(111) a. A-tsanga-ko ny lai. PASS-put up-by me the tent ‘The tent is put up by me.’ b. Voa-tsangana ny lai. PASS-put up the tent ‘The tent is put up.’ c. Tofa-tsanganany lai. PASS-put up the tent ‘The tent is put up.’
As mentioned in section 5.1, (111a-c) convey different senses; (111a) is neutral,
- 71 -
(111b) is for a perfective meaning, and (111c) describes a spontaneous event.
Turning to suffixal passives in Korean, it seems that there is a similar
phenomenon to (111), though the passive marker is identical.
(112) a. Mia-ka Inho-hanthey cap-hi-ess-ta. Mia-Nom Inho-DAT catch-PASS-PAST-DC ‘Mia was caught by Inho’ b. ku chayk-un an phal-li -n-ta. DET book-TOPIC not sell-PASS-PRES-DC ‘The book sells poorly.’ c. mwun-i cecello yel-li -ess-ta. door-NOM all by itself open-PASS-PAST-DC ‘The door opened all by itself.’
(112a) is the neutral passive. The theme role is realized at the position
of the subject, and the agent role with the dative case can be freely omitted.
There is no restriction on tense or aspectual markers as well.
(112b) may fall under the so-called middle constructions exemplified in
(113)47.
(113) a. Das Buch liest sich leicht. this book read REFL easily ‘This book reads easily.’ (German, Siewierska 1984:170) b. Cette etoffe se repasse rapidement. this fabric REFL iron rapidly ‘This fabric irons rapidly.’ (French, Fagan 1992:59) c. Boku wa nemur-are-nakat-ta. I TOPIC sleep-POTEN(PASS)-NEG-PAST ‘I could not sleep’ (Japanese, Shibatani 1985:823)
47 Jespersen (1948:350) regards the construction like (113) as ‘activo-passive use of some verbs’.
“When we say ‘his novels sell very well’, we think to some extent of the books as active themselves, as the cause of the extensive sale, while we are not thinking so much of the activity of the bookseller.”
- 72 -
Yeon (2005) claims a portion of passive sentences in Korean convey a sense of
‘potentiality,’ which is the core property of middles. Mok and Kim (2006)
argue that it is more natural to translate sentences like (112b) into English as
‘the book sells poorly’ rather than as ‘the book is not sold.’ These claims have
significance with regard to processing Korean passives, because it is not too
much to say that grammar engineering aims to generate meta-information for
machine translation. Of course, there is the opposite viewpoint, which claims
that there is no middle construction in Korean. In the following section,
contra the opposite, I will clarify my argument that there exist middles in
Korean.
(112c) denotes an event that happens spontaneously, which is much
similar to (111c). Yeon (2001:383) regards (114b), in which the patient of the
transitive verb appears as the subject of the intransitive verb, as
‘anticausative’48 . He argues ‘the anticausative construction denotes a
spontaneous process with no implied agent.’
(114) a. The child melted the ice. b. The ice melted.
Haspelmath (1993a), from a typological viewpoint, examines various forms
with regard to ‘causative/anticausative’ alternation. Though he does not treat
Korean data, passive suffixes in Korean can also play a similar role to make the
anticausative constructions. (115) shows a clear difference between neutral
passives and passives which denote a spontaneous event. Since the sentence
implicates a situation without external influence, (115a-b) sound a little strange,
as if it were a clumsy translation, while (112a), which belongs to the neutral
passives, is not bad. That is, in contrast with neutral passive sentences like
48 Other literature often calls this construction ‘ergative’ or ‘inchoative’. In order to prevent confusion, this thesis tentatively names the relevant type v-spon-pass.
- 73 -
(112a), the agent role seldom appears in this construction.49
(115) a. ?mwun-i Mia-hanthey yel-li-ess-ta. door-Nom Mia-DAT open-PASS-PAST-DC ‘(Literally) The door opened by Mia’ b. ?mwun-i Mia-ey uyhayse yel-li-ess-ta. door-Nom Mia-by open-PASS-PAST-DC ‘(Literally) The door opened by Mia’
Siewierska (1984:78) draws a sharp line between passives and anticausatives
from a cross-linguistic standpoint. The subject of passives is depicted as
‘bearing no responsibility’ for the situation or state, while that of the
anticausative construction seems to ‘bear responsibility.’
In sum, reflecting these phenomena in Korean passives, I would
introduce the three categories in (112) into the discussion of this thesis. In
order to treat them within the constraint-based framework, the type hierarchy
for suffixal passives has to be revised. (116) indicates that v-pass is divided
into three subtypes; v-neutral-pass, v-mdl-pass, and v-spon-pass.
(116)
49 It is said that there is no implicit argument in this kind of sentence, while the middle construction implies an unrealized argument.
- 74 -
The constraint on v-pass will be revised as below. Now, the procedure
of argument alternation depends on subtypes.
(117)
+−
+⇒
SSTYPE.SUFPAPASS
PASSIVE STEM
PASSIVE
-
v-tr
v-pass
(118)
⇒
1
1
SEM
)](NP[ ,NP STARG
ARG2
ARG1
INDEX
SEM
NP ,NP STARG
STEM
idat
j-
j
i
s
ji-
passv-neutral-
6.1.1 Korean Middles with Passive Suffixes
This section investigates Korean middles, which generally convey a sense of
‘potentiality.’ (119) is adapted from Levin (1993:26), and the Korean middles
that this thesis will treat are exemplified in (120c).
(119) a. The butcher cuts the meat. (active) b. The meat cuts easily. (middle)
(120) a. Inho-ka i cha-lul Mia-eykey phal-ass-ta. Inho-NOM this car-TOPIC Mia-DAT sell-PAST-DC ‘Inho sold this car to Mia.’ (active)
- 75 -
b. i cha-nun Mia-eykey phal-li-ess-ta. this car-TOPIC Mia-DAT sell-PASS-PAST-DC ‘This car was sold to Mia.’ (passive) c. i cha-nun cal phal-li-n-ta. this car-TOPIC well sell-PASS-PRES-DC ‘This car sells well.’ (middle)
As aforesaid, typical passive forms in Korean contain suffixes such as -i, -hi, -li ,
and -ki. Interestingly, the same suffixes are also used to form middles such as
(120c). The middle construction in Korean is coded as a passive form, while
English middles are active forms. Cho (1998) argues that middles are
nonexistent in Korean, and that the middle-like construction in Korean is
actually a subtype of passives. Cho’s claim notwithstanding, it is widely
known that there is a close correlation between middles and passives.50 In this
section, I argue that Korean has middles which are distinct from neutral
passives.
The previous literature on the middle construction, such as Keyser and
Roeper (1984), Bassac and Bouillon (2002), etc., has shown three major
restrictions on middles: namely syntactic, semantic, and distributional
properties.
Syntactically, it has been argued that middles imply an agent role which
cannot be projected overtly. The external argument of the predicate becomes
implicit; therefore, it cannot play a syntactic role in the sentence. From this
standpoint, the previous studies have advanced some tests for differentiating
middles from ergatives which is similar to neutral passives or v-spon-pass in
(116). The most well-known test is to insert an expression ‘all by itself’ into
the sentence. However, ‘all by itself’ insertion cannot be a perfectly reliable
test to show that there are no middles in Korean, because ‘all by itself’ can co-
50 For more information about this, see Shibatani (1985) or Kemmer (1993a).
- 76 -
occur with middles even in English, in certain cases (Chung 2002:351).
Besides, Rapoprot (1999) claims that verbs which include an ‘instrument or
manner’ component disallow ‘all by itself’ like as given in (121a), while
‘simple change-of-state’ verbs can allow it as (121b). The same goes for
Korean as exemplified in (121c-d).
(121) a. *This kind of bread cuts easily all by itself. (instrument / manner) b. These heavy windows open easily all by themselves. (simple change-of-state) c. ??i ppang-un cecello cal cal-li-n-ta. this bread-TOPIC all by itself well cut-PASS-PRES-DC ‘This bread cuts well all by itself.’ (instrument / manner) d. i chang-un cecello cal yel-li-n-ta. this window-TOPIC all by itself well open-PASS-PRES-DC ‘This window opens well all by itself.’ (simple change-of-state)
Semantically, it has been argued that middles have stative meanings not
eventive. Because of this aspectual property, middles basically cannot be
transformed into the progressive or the imperative. Cho (1998) claims the
sentence such as (122) is the evidence that the construction under discussion
does not belong to middles.
(122) i cha-nun cal phal-li-ko iss-ta. this car-TOPIC well sell-PASS-COMP be-DC ‘(Literally) This car is selling well.’
It is clear that there is an eventive aspect in (122), but I suggest that (122) also
cannot be critical evidence for the nonexistence of Korean middles. The
reason why middles cannot denote an event lies in the fact that a middle
- 77 -
construction is a generic statement. According to Steinbach (2002), a middle
verb is a kind of stage-level predicate51 which conveys the sense of a habitual
situation. If we assume all middles belong to the subset of generic expressions,
it is more important to compare a middle sentence with a generic sentence
rather than with a stative sentence. (123a) taken from Jun (1997:297) is
considered as a typical Korean generic sentence which employs a stage-level
predicate. Note that it can be transformed into the progressive, as shown in
(123b).
(123) a. kay-nun cic-nun-ta. dog-TOPIC bark-PRES-DC ‘Dogs bark.’ b. kay-nun cic-ko iss-ta. dog-TOPIC bark-COMP be-DC ‘A dog is barking.’
If middles which are responsible for generic quantification can never be
transformed into the progressive or the imperative, how can we explain the
acceptability of (123b)? Jun (1997) argues that sentences which are composed
of a stage-level predicate such as cic- ‘bark’ in (123a) have both a generic
reading and a specific reading. That is, there is an ambiguity. It is
reasonable for (123b) to be interpreted as a specific event. (123a), on the other
hand, may be read either as a specific event or as a generic sentence. Likewise,
if a verb which includes a passive suffix is transformed into the progressive or
the imperative, it is a specific predicate which denotes an event. (122) is no
more than evidence that (120c) actually has an ambiguity between middles and
passives.
51 For more information about it, see Kratzer (1995).
- 78 -
(124) i cha-nun cal phal-li-n-ta. this car-TOPIC well sell-PASS-PRES-DC a. ‘This car is sold well (by someone).’ (neutral passive) b. ‘This car sells well.’ (middle)
Distributionally, it is said that there are co-occurrence restrictions on
middles; middles need a modifier such as an adverb or a negation operator.
Korean middles also show a tendency to co-occur with such modifiers, but it’s
not always the case. However, Steinbach (2002:35) claims there are a number
of exceptions to this generalization; the exceptions are attested in Korean. On
the other hand, although it is said that middles do not select agent-oriented
adverbs such as ‘willingly’ in English, the so-called middles in Korean do allow
them, as in (125a). Fagan (1992), however, provides examples of French
middles which allow agent-oriented adverbs as presented in (126b), which is
analogous to (125a).
(125) a. i chayk-un uyyokcekulo manhi phal-li-n-ta. this book-TOPIC willingly a lot sell-PASS-PRES-DC ‘(Literally) This book sells a lot willingly.’ (Cho 1998:185) b. Ce livre se lit avec plaisir. ‘(Literally) This book reads with pleasure.’ (Fagan 1992:157)
Fagan regards sentences such as (125b) as the non-core of middles which
include an event. On the basis of his analysis, if a verb form with a passive
suffix co-occurs with an agent-oriented modifier such as (125a), I consider the
construction to be more similar to passives than middles.
Keenan (1985:253) states that the type used for passives can often be
used to derive other VPs such as middles. Since it is so hard to draw a clear
line between passives and middles formally, he suggests that the distinction can
be made on semantic grounds. In line with his proposal, I also take the
semantic properties as the core of middles. The claims along these lines are
- 79 -
grounded upon the fact that the middle-like construction is semantically
ambiguous. A Korean sentence such as (124) is ambiguous between a middle
and a passive interpretation, while an English sentence, for example ‘these
boats sink easily,’ is ambiguous between a middle and an ergative
interpretation.52 Therefore, I differentiate Korean middles from neutral
passives with semantic operators. Middles are interpreted within a generic
operator which stands for a stative situation, whereas neutral passives are bound
by an existential quantifier which denotes a specific event. The proper
representation for (120c) is as (126) adapted from Jun (1997). Although Jun
does not refer to the relationship between generics and middles, his model to
analyze generic statements is useful in representing middles as well.
(126) GENx,s[this-car(x) & C(x,s)][sell-well(x,s)] (C : contain / s : a situation argument)
With respect to the constraints on Korean middles53, I provide the
following. Morphologically, middles consist of a verb’s root plus a passive
suffix, and this stem should combine with present tense markers. Syntactically,
I suggest the agent gets defocused in middles as well as neutral passives. The
syntactic difference between them is whether the defocused agent can be
realized freely. Semantically, middles are invariably generic statements,
particularly habitual sentences. If a linguistic category which denotes a
52 This example is taken from Zubizarreta (1987:141). 53 Chung (2002) divides English middles into two subgroups, namely plain and reflexive middles. Building on his claim, I regard plain middles as the unmarked one which takes an active form and reflexive middles as the marked one which includes a special word or an affix. The unmarked type imposes rather rigid restrictions on their constructions, because if not, it is difficult to discriminate between middles and typical actives. On the other hand, the marked type which is already differentiated from actives, such as French middles (Fagan 1992), does not need such rigid restrictions. Korean middles are akin to the latter, because passive suffixes make a marked formation and play a role similar to reflexives in French, Spanish or Russian middle sentences (Shibatani 1985).
- 80 -
specific event intervenes with the construction, the construction is excluded
from middles. Distributionally, middles tend to co-occur with adverbs or
negatives, but the modifier may be deleted for pragmatic reasons.
A schema of the constraints discussed so far is presented in (127).
(127)
++
⇒
j
s
j-
j
i
s
ji-
present
v-mdl-pass
ARG2
ARG1
INDEX
SEM
NP STARG
ARG2
ARG1
INDEX
SEM
NP ,NP STARG
STEM
MIDDLE
STATIVE
TENSE
HEAD|SYN
φ1
1
The TENSE feature in the first line will be used for blocking the possibility of
combining with any other tense markers, and [STATIVE +] ensures stative
meaning of this construction. The process of argument alternation is different
from v-neutral-pass; the first argument of v-lxm cannot be projected after
alternation. Finally, φ value of ARG1 representing the external argument
becomes implicit.
- 81 -
6.1.2 A Spontaneous Event
It is a well-known fact that passive markers sometimes play the function
building a construction that denotes a spontaneous event as exemplified in
(128) taken from Shibatani (1985). According to Shibatani’s explanation, a
passive morpheme in Japnaese -(r)are or a passive morpheme in Ainu a- can be
used to denote a spontaneous event. In Spanish, the reflexive se plays the role
of a marker that implies a spontaneous event as well as the passive marker.
(128) a. Mukasi ga sinob-are-ru. old time NOM think about-SPON(PASS)-PRES ‘An old time comes (spontaneously) to mind.’ (Japanese) b. Chip a-nukar. ship SPON(PASS)-see ‘A ship is visible/seen.’ (Ainu) c. Se abrió la puerta. REFL open DET door ‘The door opened.’ (Spanish)
As mentioned before, the same phenomenon exists in Korean passives. This
section offers a solution to the phenomenon within the HPSG framework.
Chang (1995:50) says if passive suffixes attach to a transitive verb, the
transitive will be derived into an intransitive verb or a passive verb.
According to his claim, if it is possible to recognize the agent, the verb belongs
to the passives; otherwise, it is an intransitive verb. (129) is adapted from
Chang (1995:51).
(129) a. kulca-ka cal po-i-n-ta. character-NOM well see-PASS-PRES-DC ‘The character is conspicuous.’ (middle) b. malsoli-ka cal tul-li -n-ta. voice-NOM well hear-PASS-PRES-DC ‘The voice is quite audible.’ (middle)
- 82 -
c. mwun-i yel-li -ess-ta. door open-PASS-PAST-DC ‘The door got opened (by someone).’ (passive) ‘The door opened (all by itself).’ (intransitive) d. ai-ka moki-ey mwul-li -ess-ta. baby mosquito-DAT bite-PASS-PAST-DC ‘The baby was bitten by a mosquito.’ (passive)
(129a-b) share most of properties with v-mdl-pass. (129c) can be read
ambiguously. One reading is ‘someone anonymous had already opened the
door.’ The other reading is ‘by unknown reason, the door became open
without any external force.’ In contrast with (129c), (129d) where the agent
appears has no ambiguity. In the previous section, I claimed that middle
sentences in Korean, such as (120c) or (129a-b), can be interpreted
ambiguously. Just like Korean middles, some passive forms that may denote a
spontaneous event also tend to have an ambiguity.
The constraints on this type will be summarized as follows: Syntactically,
there is no implicit argument, unlike middles; therefore, the agent role cannot
appear after argument alternation takes place. Semantically these
constructions usually have an eventive aspect, and finally it seems that there is
no morphological or distributional restriction on this type. Consequently,
(130) stands for the constraints on v-spon-pass. Since the external argument is
deleted after argument alternation, ARG1 will not be coded in SEM.
- 83 -
(130)
−⇒
j
s
j-
j
i
s
ji-
sv-spon-pas
ARG2
INDEX SEM
NP STARG
ARG2
ARG1
INDEX
SEM
NP ,NP STARG
STEM
STATIVE
6.2 Passive-like Forms
In (20), I defined the scope of canonical passives in Korean. Normally, only
transitive verbs can be passivized, and there is a corresponding active form for a
passive form. All of the constructions dealt with so far are compatible with
these criteria. However, there are small but considerable exceptions to this
generalization. Constructions that this chapter covers stand out of the normal
scope of passives.
6.2.1 Passive Forms of Intransitive Verbs
Investigating phenomena regarding passives in world languages from a cross-
linguistic viewpoint, it is not a surprising fact that an intransitive verb can be
passivized. (131) is a case in which an intransitive verb is transformed into a
passive in Latin (Keenan 1985:250). The same goes for Korean as presented
in (132), though it is unproductive.
- 84 -
(131) a. currere ‘to run’ b. curritur ‘(it) is run, running is being done’
(132) a. Mia-nun col-ass-ta Mia-TOPIC drowse-PAST-DC ‘Mia drowsed.’ b. Mia-nun col-li -ess-ta Mia-TOPIC drowse-PASS-PAST-DC ‘Mia grew drowsy.’
How can this phenomenon be taken into account within the HPSG
framework? One of the conceivable claims is that col-li- ‘grow drowsy’ in
(132b) is a passive form of a predicate that has been already transformed into a
causative. This claim seems relatively attractive, but there are two vulnerable
spots. First, this approach ignores the fact that passive suffixes and causative
suffixes share the same slot in verbal morphology. In type hierarchy (46) for
verbal morphology, v-pass and v-caus stand in the relation of sisterhood;
therefore they cannot appear at once. Secondly, Im and Chang (1995:423)
criticize this approach; suffixal causatives are seldom derived into suffixal
passives.54
Chang (1995), who claims that suffixal passives without an agent may
belong to the intransitive verbs, offers an alternative to the above. That is, I
would regard col-li- ‘grow drowsy’ in (132b) as a root, not an inflectional form.
On this assumption, a lexicon colli- ‘grow drowsy’ can be written in the
dictionary as an element of the set of intransitive verbs.
Let me make a comparison between colli- and the so-called ‘deponent
verbs’ exemplified in (133) from Kiparsky (2005:119). (133a-b) is an ordinary
active/passive pair in Latin, and (133c) is a typical example of a deponent verb. 54 Im and Chang (1995) regard this kind of verbs as irregular ones. However, Haspelmath (1990:46) argues that ‘passives from causatives’ can be a source for passive from a typological view, though it is not well-known.
- 85 -
(133) a. laudat ‘praises’ b. laudātur ‘is praised’ c. hortātur ‘exhorts
Deponent verbs refer to ‘a class of syntactically active and morphologically
passive verbs,’ which is much similar to colli- ‘grow drowsy’.55 Kiparsky,
furthermore, claims that ‘deponent verbs are inherently specified as [PASSIVE
+].’ In line with his argument, I assume that colli- also has [PASSIVE +]
information inherently. Besides, colli- cannot be passivized once again. This
thesis, therefore, suggests colli- belongs to the inherent passive verbs, such as
mac- ‘be hit’.
However, there is a difference in usage between colli- ‘grow drowsy’ and
mac- ‘be hit’. Whereas the appearance of an agent role has no effect on
acceptability in sentences with mac-, an agent role rarely co-occurs with colli-.
(132b) implies that ‘nobody makes Mia feel sleepy,’ which is similar to (134)
which denotes a spontaneous event in Russian.
(134) On utomil+sja he exhausted+MM ‘He grew weary.’ (Kemmer 1993b:27)
That is, colli- ‘grow drowsy’ shares some properties with v-spon-pass. I
tentatively suggest the type inherent-passive-v2 in order to represent the lexeme
colli-.
55 As presented in section 3.2, Kemmer (1993a) claims that there are deponent verbs in all languages with passives.
- 86 -
(135)
−−
−+
i
ygrow-drows
s
nom-
-
ass-vinherent-p
colli
i
ARG1
RELN RESTR
INDEX
SEM
]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
STATIVE
PASSIVE
2
It may be controversial that this representation cannot capture the
relationship between col- ‘drowse’ and colli- ‘grow drowsy’. Nevertheless,
since the number of passive forms of intransitive verbs is quite small and this
method is based on previous solutions to canonical passives, I believe my
approach is well-balanced.
6.2.2 Obligatory Passives
It is said that active/passive pairs have close relations with each other because
they share a truth condition. It is very hard to find a case in which an active
sentence, in spite of a false value of its corresponding passive sentence, has a
truth value, and vice versa. However, there are some exceptional cases in
which there exists an asymmetry between actives and passives
An obligatory passive verb refers to a passive form without its
corresponding active form. That means the verb should always be used in the
form of passives. Levin (1993:107) takes an instance in English; if ‘rumor’ is
used as a verb, the construction will be ‘it is rumored that’ in most cases.
- 87 -
(136) a. It is rumored that he left town. b. *They rumor that he left town.
In Korean, some passive forms have no corresponding active forms. (137) is
adapted from Im and Chang (1995:423).
(137) a. ??cong-i wu-n-ta. bell-NOM ring-PRES-DC ‘The bell rings.’ b. cong-i wul-li -n-ta. bell-NOM ring-PASS-PRES-DC ‘The bell rings.’
I think this wul-li- ‘ring’ in (137b) is much similar to col-li- ‘grow drowsy’ in
(132b). Since the same solution is applicable to obligatory passives in Korean,
I regard wulli- ‘ring’ as one of the inherent passive verbs. Just as the previous
cases, such as colli- ‘grow drowsy’ in (135), the number of these verbs is very
small. Therefore, this method won’t increase lexical redundancy by much.
(138)
−−
−+
i
s
nom-
-
ass-vinherent-p
wulli
i
ARG1
INDEX SEM
]NP[ STARG
AUXPASS
SUFPASS TYPEPASS
STATIVE
PASSIVE
2
- 88 -
6.3 Double Passives
Im (1998:339) proposes the so-called ‘single passive constraint’ as one of the
constraints on Korean passives. That means passivization cannot take place
twice in a sentence; therefore Im criticizes the term ‘double passives.’ Those
who study the traditional prescriptive grammar which disallows ‘double
passives’ probably regard (139c) as incorrect usage. However, some verbs,
especially ic- ‘forget’, are more frequently used in the form of double passives.
(139) a. ku saken-i ic-hi-ess-ta. DET incident-NOM forget-PASS-PAST-DC ‘The incident is forgotten.’ b. ku saken-i ic-e ci-ess-ta. DET incident-NOM forget-COMP AUXPASS-PAST-DC ‘The incident is forgotten.’ c. ku saken-i ic-hi-e ci-ess-ta. DET incident-NOM forget-PASS-COMP AUX -PAST-DC ‘The incident has been forgotten.’
In (139) ic- is followed by a passive suffix hi-, and then ic-hi- ‘be forgotten’
combines with ‘e ci-’ which makes an auxiliary passive formation.
To begin with, let us find out which verbs are usually used in the form of
double passives. Song (2007) makes use of the Expected Likelihood Estimator
(henceforth ELE) in order to calculate the weight of co-occurrence relationship
between adjectives and general nouns. This thesis also applies ELE to
measurement, using the Sejong POS-tagged corpora. In (140), C is short for
the ‘corresponding frequency,’ N means the whole frequency of the verbs, and B
stands for the total frequency of all verbs.
(140)
BN
wwCwwP n
n ×+
+…=…
5.0
5.0)()( 1
1
- 89 -
I measured the weight of co-occurrence between suffixal passive forms and a
passive auxiliary ci-, for example ic-hi- and ci-, using ELE formula in (140).
Table (5) contains the top 5 of the result.
Table (5) Verb Freq(A) -e ci-(B) ELE Prop (B/A)
ic-hi- ‘be forgotten’ 254 253 0.0240877993 99.61% ssu-i- ‘be used/written’ 336 27 0.0214501226 67.56%
cca-i- ‘be framed’ 186 159 0.0152543994 85.48%
po-i- ‘be seen’ 3725 187 0.0133976420 5.02% pwul-li- ‘be called’ 349 117 0.0110650720 33.52%
Table (5) implies that some verbs, such as ic-, are mostly used in the form of
double passives, in reality. I claim that this phenomenon should be taken into
consideration.
Since ic- belongs to (p-3) type according to classification presented in
(39), it is clear that both (139a) and (139b) are acceptable and regular forms in
Korean. The problem that I would like to raise is how to explain (139c).
If I draw a conclusion in advance, (139c) differs from (139a-b) in some
respects. As aforesaid in section 4.3.3, (139c) conveys a sort of resultative
meaning, which is similar to perfective aspect in English. Haspelmath
(1993b) introduces the so-called ‘resultative particles’ which are derived from
passive markers. For example, let us think over an English expression ‘an
escaped prisoner.’ A prisoner had already escaped from a prison, and as a
result of the event, he or she exists there no longer. In this case, ‘-ed’ plays the
role of a resultative particle. An auxiliary ci- in (139c), likewise, seems to
function like a resultative particle. That is, ci- in (139c) is subtly different
from the ordinary passive auxiliary ci- represented in (61).
In order to argue that (139c) is different from (139a-b) with respect to
aspectual properties more clearly, let me make a comparison between the
- 90 -
sentences in (141).
(141) a. ?ku saken-i ic-hi-ko iss-ta. DET incident-NOM forget-PASS-COMP be-DC ‘(Literally) The incident is being forgotten.’ b. ?ku saken-i ic-e ci-ko iss-ta. DET incident-NOM forget-COMP AUXPASS-COMP be-DC ‘(Literally) The incident is being forgotten.’ c. ku saken-i ic-hi-e ci-ko iss-ta. DET incident-NOM forget-PASS-COMP AUX-COMP be-DC ‘(Literally) The incident has been being forgotten.’
Sentences that have stative properties tend hardly to be transformed into
progressives, while sentences that denote an apparent event do not (Sag
1973:88). If a predicate is transformed into a passive, the eventive meaning of
the predicate normally weakens. I suppose that it is because the agent gets
defocused after passivization (Shibatani 1985). Hence, the progressive of
passives tends to become a little awkward in comparison with that of their
corresponding actives, in Korean.
(142) a. Inho-ka Mia-lul cap-ko iss-ta. Inho-NOM Mia-ACC catch -COMP be-DC ‘Inho is catching Mia.’ b. ?Mia-ka Inho-hanthey cap-hi-ko iss-ta. Mia-NOM Inho-DAT catch-PASS-COMP be-DC ‘Mia is being caught by Inho.’
In this context, the sentences in (141) show a difference in aspectual property
between single passives and double passives. (141c) sounds better than the
others according to my intuition. Besides, constructions like (141c) appear
several times in the Sejong POS-tagged corpora, whereas those like the others
do not.
- 91 -
I suggest that the so-called double passives are differentiated from single
passives, and it is likely to be the reason why some verbs such as ic- are usually
used in the form of double passives. Furthermore, I assume that ci- in (139c)
or (141c) is a distinct lexicon from a passive auxiliary ci- represented in (61),
because they have some distinguishable properties from each other as shown in
(141b-c).
Unless this thesis stands on the above assumption, the constraint on a
passive auxiliary ci- needs to be revised. It is because (61) assumes that a
passive auxiliary ci- can take only [PASSIVE −] verbs as its complement.
Therefore, a construction like (139c), whose main verb has been already
specified as [PASSIVE +], cannot be handled as hd-lex-pass-ex. Since ci- in
(139c) does not lead argument alternation, ic-hi-e ci- in (139c) should be dealt
with by other constraints; in particular, hd-lex-nonalt-lex. (143) is the lexical
representation of ci- in double passive constructions, and (144) is the lexical
representation for hd-lex-nonalt-lex.
(143)
−
+
11
SUBJ
VFORM ],NP[ STARG
PASSIVE
AUX
aenom-
aux
ci
(144) [ ]Vv-compalt-exhd-lex-non
, COMPS COMPS
→
11
Despite the term ‘double passive,’ this method adheres to the ‘single passive
constraint’ of Im (1998), in consequence.
- 92 -
7. Conclusion
The main purpose of this thesis is to contribute to the development of the
Korean computational grammar, such as KRG. This thesis built up constraints
on passives in Korean within the HPSG framework on the basis of more
articulated analyses, and provided fine-grained type hierarchies for passives in
Korean. The type hierarchies proposed in this thesis are grounded on the
classification of Korean passives; suffixal passives, auxiliary passives, inherent
passives, and passive light verb constructions.
In this thesis, I proposed comprehensive type hierarchies for verbs or
verbal nouns with respect to passivization. I classified verbs into five subtypes
with reference to passivization. I modified the verbal morphology of Kim and
Yang (2006) in order to treat suffixal passives; in particular, the v-alt-stem was
introduced into the verbal morphological hierarchy. For auxiliary passives, I
introduced hd-lex-pass-ex into the syntactic hierarchy. Turning to verbal
nouns, I proposed a classification of verbal nouns regarding which passive light
verbs they can combine with. A type hierarchy for passive light verbs was
proposed in this study as well.
In addition to the above canonical passives, the related constructions
were discussed, too; for example, middles, spontaneous events, obligatory
passives, etc. This discussion aims to make more comprehensive analyses of
Korean passives and passive-like constructions.
I have implemented type hierarchies and constraints for passives in the
LKB system to check their computational feasibility. All sample sentences in
this thesis were tested in LKB, and it may merit further research to inspect a
larger range of data in the system. An important part of the implementation
will be given in the Appendix.
Finally, I would like to point out the methodology taken in this study, that
- 93 -
is, to make use of language resources available in an extensive and
comprehensive way. I provided full details of the methodology of the data
compilation for the research in this thesis. I believe this kind of descriptive
and inductive approach complements the more theoretically oriented
approaches. I also believe that it is an efficient way to figure out the nature of
language.
- 94 -
References Ackerman, Farrell and Webelhuth, Gert. 1998. A Theory of Predicates. Stanford: CSLI
Publications. Alsina, Alex, Joan Bresnan, and Peter Sells (eds.). 1996. Complex Predicates.
Stanford: CSLI Publications. Bassac, Christian and Pierrette Bouillon. 2002. Middle Transitive Alternations in
English: A Generative Lexicon Approach. In Many Morphologies, ed. Paul Boucher, 29~47. Somerville, MA: Cascadilla Press.
Bresnan, Joan. 1982. The Passive in Lexical Theory. In The Mental Representation of Grammatical Relations, ed. Joan Bresnan, 3-86. Massachusetts: MIT Press.
Chae, Hee-Rahk. 1996. Properties of ha- and Light Predicate Constructions. Language Research 32:409-476. (In Korean).
Chae, Hee-Rahk. 2003. Passive Light Verb Constructions in Korean. Harvard Studies in Korean Linguistics 10.
Chang, Suk-Jin. 1995. Information-based Korean Grammar. Seoul: Hanshin Publishing. (In Korean).
Cho, Young-Mee Yu and Peter Sells. 1995. A Lexical Account of Inflectional Suffixes in Korean. Journal of East Asian Linguistics 4:119-174.
Cho, Sungdai. 1998. Passive and Middle Constructions in Korean. Korean Linguistics 9:151~189.
Choi, Incheol and Stephen Wechsler. 2002. Mixed Categories and Argument Transfer in the Korean Light Verb Construction. Paper presented at The 8th International Conference on Head-Driven Phrase Structure Grammar, Norwegian University of Science and Technology, Trondheim, Norway.
Chomsky, Noam. 1995. The Minimalist Program. Massachusetts: MIT Press. Chung, Taegoo. 2002. The Reflexive Middles in English. Studies in Generative
Grammar 12:339~371. (In Korean). Copestake, Ann. 2002. Implementing Typed Feature Structure Grammars. Stanford:
CSLI Publications. Copestake, Ann, Dan Flickinger, Carl Pollard, and Ivan A. Sag. 2005. Minimal
Recursion Semantics: An Introduction. Research on Language and Computation 3:281-332.
Fagan, Sarah M. 1992. The Syntax and Semantics of Middle Constructions: A Study with Special Reference to German. Cambridge: Cambridge University Press.
Fromkin, Victoria, Robert Rodman, and Nina Hyams. 2003. An Introduction to Language 7th edition. Boston: Heinle.
Givón, Talmy. 1991. Isomorphism in the Grammatical Code: Cognitive and Biological
- 95 -
Considerations. Studies in Language 15:85-114. Haspelmath, Martin. 1990. The Grammaticalization of Passive Morphology. Studies in
Language 14:27~71. Haspelmath, Martin. 1993a. More on the Typology of Inchoative/Causative Verb
Alternations. In Causatives and Transitivity, eds. Bernard Comrie and Maria Polinsky, 87~120: John Benjamin Publishing Co.
Haspelmath, Martin. 1993b. Passive Participles Aross Languages. In Voice: Form and Function, eds. Barbara Fox and Paul J. Hopper, 151-177. Amsterdam: John Benjamins Publishing Company.
Im, Hong-Pin and Sowon Chang. 1995. The Theory of Korean Grammar [국어문법론] . Seoul: Korea National Open University Press. (In Korean).
Im, Hong-Pin. 1998. The Deep Structure of Korean Grammar [국어 문법의 심층] 3. Seoul: Thayhaksa. (In Korean).
Jespersen, Otto. 1948. A Modern English Grammar on Historical Principles, Part 3 Syntax. London: George Allen & Unwin LTD.
Jun, Youngchul. 1997. Types of Korean Generic Sentences. Eoneohag 21:289~303. (In Korean).
Kang, Beom-Mo. 2001. Semantic Structures of Predicative Nouns. Eoneohag 31:3-29. (In Korean).
Kang, Myung-Yoon. 1997. Bare Phrase Approach to Korean Morphological Causative/Passive Construction. Language Research 33:79-100. (In Korean).
Keenan, Edward L. 1985. Passive in the world's languages. In Language typology and syntactic description, ed. Timothy Shopen, 243-281. Cambridge: Cambridge University Press.
Kemmer, Suzanne. 1993a. Middle Voice, Transitivity, and the Elaboration of Events. In Voice: Form and Function, eds. Barbara Fox and Paul J. Hopper, 179~230. Amsterdam: John Benjamins Publishing Company.
Kemmer, Suzanne. 1993b. The Middle Voice. Amsterdam: John Benjamins Publishing Company.
Keyser, Samuel Jay and Thomas Roeper. 1984. On the Middle and Ergative Constructions in English. Linguistic Inquiry 15:381~416.
Kim, Seok-Duk. 1992. Korean Morphology [우리말 형태론] . Seoul: Thap Press. (In Korean).
Kim, Soowon and Joan Maling. 1998. Resultative: English vs. Korean. Japanese/Korean Linguistics 7:363-380.
Kim, Chung Ja. 2002. Study on the Antecedent of the Passive Construction with Transitive Verbs. Comparative Korean Studies 10:65-95. (In Korean).
Kim, Jong-Bok. 2004. Korean Phrase Structure Grammar. Seoul: Hankwuk
- 96 -
Publishing. (In Korean). Kim, Jong-Bok. 2005. Korean Passive Constructions: A Nonderivational, Constraint-
Based Perspective. Korean Linguistics 26:67-92. (In Korean). Kim, Jong-Bok and Jaehyung Yang. 2004. A Constraint-based Approach to Korean
Auxiliary Contructions and its Computational Implementation. Language Research 40:195-226. (In Korean).
Kim, Jong-Bok, and Jaehyung Yang. 2006. Developing a Korean Resource Grammar and Implementing It into the Linguistic Knowledge Building System. The Journal of Linguistic Science 37:23-67. (In Korean).
Kim, Jong-Bok, Jaehyung Yang, and Incheol Choi. 2004. Capturing and Parsing the Mixed Properties of Light Verb Constructions in a Typed Feature Structure Grammar. Paper presented at The 18th Pacific Asia Conference on Language, Information, and Computation, Waseda University.
Kiparsky, Paul. 2005. Blocking and Periphrasis in Inflectional Paradigms. Yearbook of Morphology 2004:113-136.
Kratzer, Angelika. 1995. Stage-level and Individual-level Predicates. In The Generic Book, eds. Greg N. Carlson and Francis J. Pelletier, 125-175. Chicago: The University of Chicago Press.
Kwon, Jae-il. 1993. A Historical Study of the Passive in Korean. Eoneohag 15:25-43. (In Korean).
Lee, Sang-Oak. 1999. A Study of Causative and Passive Contstruction in Korean. Seoul: Jipmoondang. (In Korean).
Lee, Jung-Tag. 2001. A Study of the Passive Expressions Which have One of the Verbs 'toeda, patta, tanghada'. Hangeul 251:143-166. (In Korean).
Lee, Jung-Tag. 2004. A Study of Passives in Contemporary Korean [현대 국어 피동 연구] . Seoul: Pakijeng. (In Korean).
Lee, Junkyu and Chungmin Lee. 2002. Korean Resultative Constructions. Paper presented at The 9th International Conference on Head-Driven Phrase Structure Grammar, Kyung Hee University.
Lee, Nam-geun. 2005. A Constraint-based Analysis on English and Korean Passive Construction. The Linguistic Association of Korean Journal 13:227-247. (In Korean).
Levin, Beth. 1993. English Verb Classes and Alternations: a Preliminary Investigation. Chicago: University Of Chicago Press.
Mok, Jung-soo and Kim, Yeong-jung. 2006. A Structure of Passive Constructions in Korean and their meaning 'Potential'. Humanistic Studies in Language 8:369~387. (In Korean).
Müller, Stefan. 2000. The Passive as a Lexical Rule. Paper presented at The 7th
- 97 -
International Conference on Head-Driven Phrase Structure Grammar Berkeley, University of California.
Müller, Stefan. 2003. Object-To-Subject-Raising and Lexical Rule: An Analysis of the German Passive. Paper presented at The 10th International Conference on Head-Driven Phrase Structure Grammar, Michigan State University.
Müller, Stefan and Ivan A. Sag. 2007. Introduction to HPSG. In Lecture Slide of 2007 LSA Linguistic Institute, Stanford University.
Murphy, Patrick M. 2004. Passive Prototypes, Topicality, and Conceptual Space, Dept. of Linguistics, University of North Carolina at Chapel Hill: Ph. D.
Park, Hyo Myong. 2004. The Theory of Head-Driven Phrase Structure Grammar [핵어문법론 II] . Gwangju: Chonnam National University Press. (In Korean).
Pollard, Carl and Ivan A. Sag. 1994. Head-driven Phrase Structure Grammar. Chicago The University of Chicago Press.
Poser, William J. 1992. Blocking of Phrasal Constructions by Lexical Items. In Lexical matters, eds. Ivan A. Sag and Anna Scabolcsi. Stanford: Stanford University Press.
Rapoprot, Tova R. 1999. The English Middles and Agentivity. Linguistic Inquiry 30:147~155.
Ryu, Byong-Rae. 2006. A Fragment of Constraint-based Korean Grammar and Its Verification based on the Linguistic Knowledge Building System - with a Special Focus on Light Verb Constructions. Studies in Humanities 33:145-176. (In Korean).
Sag, Ivan A. 1973. On the State of Progress on progressives and statives In New Ways of Analyzing Variation in English, eds. Charles-James N. Bailey and Roger W. Shuy. Washington, DC: Georgetown University Press.
Sag, Ivan A. 2007. Locality in Sign-Based Construction Grammar. Paper presented at The 14th International Conference on Head-Driven Phrase Structure Grammar, Stanford University.
Sag, Ivan A and Thomas Wasow. 1999. Syntactic Theory. Stanford: CSLI Publications. Sag, Ivan A., Thomas Wasow, and Emily M. Bender. 2003. Syntactic Theory: A formal
introduction, 2nd Edition. Stanford: CSLI Publications. Saussure, Ferdinand de. 1916. Grundfragen der allgemeinen Sprachwissenschaft.
Berlin: Walter de Gruyter & Co. Sells, Peter. 1995. Korean and Japanese Morphology from a Lexical Perspective.
Linguistic Inquiry 26:277-325. Shibatani, Masayoshi. 1985. Passives and Related Constructions: A Prototype Analysis.
Language 61:821~848. Siewierska, Anna. 1984. The Passive : A Comparative Linguistic Analysis. London:
- 98 -
Croom Helm. Sohn, Ho-Min. 1999. The Korean Language. Cambridge: Cambridge University Press. Song, Sanghoun. 2007. The Retrieval of the Co-occurring Relation between Adjectives
and General Nouns on the Basis of Document Models. KLing 1:163-190. (In Korean).
Song, Sanghoun and Jae-Woong Choe. 2007. Type Hierarchies for Passive Forms in Korean. Paper presented at The 14th International Conference on Head-Driven Phrase Structure Grammar, Stanford University. (http://cslipublications.stanford.edu/HPSG/8/song-choe.pdf)
Steinbach, Markus. 2002. Middle Voice: A Comparative Study in the Syntax-Semantics Interface of German. Amsterdam: John Benjamins Publishing Company.
Sung, Kwang Su. 1976. Korean Pseudopassives [국어 의사피동문]. Collected Papers of Kwangdong University 4. (In Korean).
Yang, Dong-Whee. 1979. Passive and Causative Verbs in Korean. Hangeul 166:189-206. (In Korean).
Yeon, Jaehoon. 2001. Transitivity Alternation and Neutral-verbs in Korean. Bulletin of the School of Oriental and African Studies 64:381~391.
Yeon, Jaehoon. 2002. The Causative-passive Ambiguity and the Notion of "Contiguity" as a Crucial Factor in Explaining the Retained-object Passive Constructions. Eoneohag 32:197-221.
Yeon, Jaehoon. 2005. The Typology of Passives. In Semantic Structure, Syntactic Structure, and Beyond, 581-605. Seoul: Hankwuk Publishing. (In Korean).
Zubizarreta, Maria Luisa. 1987. Levels of Representation in the Lexicon and in the Syntax. Dordrecht, Holland: Foris.
- 99 -
v-tr-p1
가로막다 갈다 change 감다 wash 감다 wind 걷다 clear off 걷다 collect 걸다 그을다 긁다 까다 깨다 꼬집다 꼽다 꽂다 꾀다 꿇다 끌다 내다보다 내려다보다 내리깔다 내몰다 내보다 내쫓다 누르다 닫다 달다 담그다 둘러싸다 뒤덮다 뒤섞다 뒤쫓다 뒤틀다 뒤흔들다 듣다 들볶다 들여다보다 떠밀다 떨다 tremble 뜨다 awake
뜨다 detach 뜨다 float 막다 맞물다 매다 weed 매달다 먹다 몰다 물다 밀다 박다 밟다 벗다 부딪다 부르다 불다 붙들다 비꼬다 빨다 launder 빨다 suck 빼앗다 사로잡다 싣다 싸다 쓰다 use 쓸다 씹다 안다 억누르다 업다 에다 여닫다 열다 엿보다 옥죄다 잠그다 sink 잡다 적다 접다 조르다 tease 집다
짓누르다 짓밟다 짜다 쪼다 찌르다 찍다 chop 차다 처박다 쳐다보다 치다 trap 트다 틀다 파묻다 팔다 헐다 휘몰다 휩싸다 휩쓸다 v-tr-p2
가꾸다 가다듬다 가로지르다 가로채다 가르치다 가리다 가하다 간추리다 감추다 감하다 갖추다 갚다 개다 개키다 갸웃거리다 거두다 거르다 건지다 걸치다 게우다
겨누다 겪다 겹치다 고치다 고하다 곤두박질치다 곱하다 구기다 구하다 그리다 그만두다 글썽거리다 글썽이다 긁적거리다 긋다 기다리다 기대다 기르다 기리다 길들다 깁다 까뒤집다 깨닫다 깨뜨리다 깨우치다 깨치다 꺼내다 꺼리다 꾀하다 꾸다 꾸리다 꾸미다 꿰뚫다 꿰매다 끄다 끄덕이다 끌어내다 끌어내리다 끌어당기다 끌어올리다 끼얹다
끼우다 끼치다 나누다 나르다 나타내다 낚다 낳다 내던지다 내동댕이치다 내디디다 내려놓다 내려뜨리다 내리다 내리치다 내맡기다 내뱉다 내버리다 내보내다 내뿜다 내세우다 내지르다 내치다 내팽개치다 넣다 논하다 놓치다 누그러뜨리다 눈여기다 뉘우치다 느끼다 늘어뜨리다 다듬다 다루다 다리다 다스리다 다지다 다투다 달구다 달래다 닳다 당기다
Appendix I (Classification)
- 100 -
대다 더듬다 던지다 덧나다 덧대다 덧바르다 덧보태다 데우다 데치다 도리다 돌려보내다 돌보다 돌이키다 되새기다 두근거리다 두다 두들기다 두르다 둘러치다 뒤바꾸다 뒤지다 들뜨다 들먹거리다 들먹이다 들여보내다 들추다 들치다 디밀다 따돌리다 따르다 떠넘기다 떠받들다 떨구다 떨치다 뜯어고치다 마치다 만나다 만들다 만지다 만지작거리다 말하다 망가뜨리다 망설이다 망치다 망하다 맞바꾸다
맞추다 맡다 매기다 맹글다 머금다 메꾸다 멸하다 모시다 몰아가다 몰아내다 몰아붙이다 몰아치다 무너뜨리다 문지르다 뭉뚱그리다 미루다 믿다 밀어붙이다 밀치다 바꾸다 바치다 받들다 받아들이다 받치다 발하다 배우다 뱉다 버리다 버무리다 버티다 벌이다 범하다 베풀다 보내다 보태다 봉하다 부릅뜨다 부리다 부수다 부추기다 부치다 부풀다 분지르다 불러일으키다 불어넣다 불타다
붓다 비비다 빌리다 빚다 빠개다 빻다 빼내다 빼돌리다 뻗치다 뿌리다 뿜다 삐다 살펴보다 살피다 삶다 삼키다 새기다 설치다 섬기다 솎다 솟구다 수놓다 스치다 싸매다 썰다 쏟다 씰룩거리다 아로새기다 아물다 아우르다 악물다 알다 앞당기다 어지르다 얻다 얼버무리다 업신여기다 엎다 엎드리다 엎지르다 여기다 여미다 연잇다 오그리다 오리다 옮다
외치다 움추리다 움키다 위하다 으깨다 읊다 이기다 이다 이루다 일구다 일으키다 일컫다 잃다 잇다 잇대다 잊어버리다 잡아당기다 저물다 저미다 저지르다 적시다 전하다 점치다 젓다 제끼다 제하다 조이다 죄다 주다 쥐다 지어내다 지지다 지키다 지피다 집어넣다 집어던지다 짓다 짓이기다 짜내다 째다 쪼개다 찌그리다 찌푸리다 찔끔거리다 찢어발기다 찧다
차리다 참다 찾다 찾아내다 채다 처넣다 추다 추리다 취하다 치루다 치르다 치켜들다 치키다 칠하다 칭하다 켜다 타다 택하다 터뜨리다 토하다 퉁기다 튕기다 파다 파헤치다 팽개치다 퍼붓다 펼치다 평하다 포개다 풀어놓다 풀어헤치다 풍기다 피하다 할퀴다 합치다 합하다 행하다 허물다 헝클다 헤다 헤아리다 휘갈기다 휘청거리다 흘리다 흥하다 흩다
- 101 -
흩뿌리다 v-tr-p3
가두다 가르다 갈다 grind 감싸다 깎다 깔다 꺾다 꼬다 꿰다 끊다 끼다 내밀다 널다 놓다 다물다 닦다 담다 덜다 덮다 뒤집다 들다 떨다 take off 떼다 뚫다 뜯다 말다 매다 tie 맺다 메다 모으다 묶다 묻다 바로잡다 받다 벌다 베다 보다 볶다 붙잡다 비추다 비틀다 뻗다
뽑다 섞다 손꼽다 심다 쌓다 쏘다 쓰다 write 씻다 얹다 얽다 엮다 이끌다 읽다 잊다 자르다 잠그다 lock 조르다 tighten 지르다 찍다 stamp 찢다 치다 strike 털다 펴다 풀다 휘감다 휘두르다 흔들다 vn-tr-plv1
감시 강요 공제 무시 방해 배척 부여 비난 비판 설득 세뇌 소환 손상 시험 신고 억압
외면 요구 유린 이송 지명 지적 징계 착취 처단 처벌 추궁 취급 침해 탄압 vn-tr-plv2
감동 감수 감화 검증 격려 결제 고취 공급 공천 구형 긍정 기대 기록 기증 기탁 낙찰 납품 단죄 대여 대접 대출 도장 면제 면책 명령 문책 반환 발급 발령
발부 배급 배달 배당 배분 배속 배정 배치 보고 보급 보상 보장 보증 보충 부과 부양 부축 분급 분배 분할 불신 불하 사면 산출 상납 상속 서비스 서품 선고 선사 선택 선호 섭외 소개 속죄 송금 송달 송치 송환 수여 수임 숭배 숭상 숭앙 승인 시사
신뢰 신임 신청 실습 안내 암시 약속 양도 양보 연결 연기 연락 예고 예방 예시 예약 예언 예우 예찬 오인 오해 요청 용서 우대 우송 원조 위로 위임 위탁 유산 유전 융자 의뢰 의심 이식 이양 이전 이해 인가 인계 인수 인식 인정 임대 입증 재촉
- 102 -
적용 전달 전도 전수 점지 접수 제공 제시 조달 조련 조명 조사 조정 조화 존중 주목 주입 증명 증여 증원 증폭 지급 지목 지불 지시 지원 지정 지지 진단 진상 차별 찬양 창조 책정 처방 초대 초청 촉망 추앙 추인 추정 추천 축복 측정 치료 치유
칭송 타진 통고 통제 통치 투여 판단 판정 하달 하사 할애 허가 허락 허용 호명 호출 확인 환불 환수 후원 훈련 훈방 vn-tr-plv3
감금 강탈 거부 거세 거절 검색 격추 결박 금지 기각 기피 난타 날치기 납치 농락 단절 도굴 도태 독살 독점 매도
매수 매장 매혹 멸망 몰수 묵살 박탈 발견 배제 변경 변신 부정 분해 비교 사살 사역 사육 삭제 살육 살해 상실 소모 속박 수용 숙청 시해 실연 악용 암살 압도 압류 압수 애무 약탈 억류 연행 유괴 유배 유실 이용 잠식 저지 저해 적출 전멸 절단
점령 점유 정리 정복 정지 제거 제명 제소 조작 조종 종속 좌절 중단 중지 진압 질식 집행 징발 징수 징용 징집 차단 참수 처분 처형 철거 체포 촉구 총살 추방 축출 침탈 타도 탈취 투옥 파악 폐간 폭파 피격 피습 학살 함락 해고 해산 해임 해제
해직 해체 현혹 회수 회유 훼손 흡수 희롱 희생 vn-tr-plv4
경멸 멸시 수술 정학 추징 침략
vn-tr-plv5
가결 가늠 가동 가량 가미 가산 가속 가식 가열 가중 가창 각색 각성 각인 간과 간주 간직 간택 갈무리 감각 감별 감산 감상 감소
- 103 -
감안 감액 감지 감축 감형 강구 강등 강변 강의 강조 강행 강화 개간 개관 개량 개명 개문 개발 개방 개봉 개선 개설 개시 개악 개원 개입 개작 개장 개정 개조 개진 개최 개축 개칭 개통 개편 개폐 개표 개화 개회 갱신 거래 거론 거명 거행 건국
건립 건설 건의 건조 건축 검거 검출 검토 게시 게양 게재 격리 격침 격파 격하 견제 견지 결단 결부 결빙 결성 결속 결여 결의 결집 결핍 결합 결행 경감 경계 경과 경도 경매 경시 경질 계류 계발 계상 계속 계시 고사 고수 고시 고안 고양 고자질
고증 고집 고착 고창 곡해 공경 공략 공모 공수 공언 공여 공지 과세 과장 과점 관류 관용 관찰 광고 괴멸 교란 교사 구매 구입 구출 구획 굴착 권장 궤멸 규합 근절 급조 급파 기도 기부 기약 기용 기입 기재 기화 긴장 긴축 낙태 날조 남발 남용
남파 남획 납득 납본 납부 낭독 낭비 낭송 내장 내포 냉각 냉동 노래 노출 노화 녹음 녹취 녹화 논평 농축 누락 누설 누적 누진 누출 단련 단선 단속 단수 단장 단정 단축 단행 달성 담당 답습 당선 대결 대답 대령 대립 대변 대별 대비 대신 대응
대입 대조 대체 대치 대표 대필 도금 도모 도발 도배 도살 도색 도입 도출 독립 독파 돌파 동결 동반 동봉 동요 동원 되풀이 둔갑 등기 등록 등분 등용 마감 마련 마무리 마취 만족 만회 말살 망각 매각 매개 매립 매매 매몰 매설 매입 메모 면직 멸균
- 104 -
명명 명시 명중 모금 모방 모색 모의 모조 모집 묘사 무마 묵인 물갈이 미혹 미화 밀반출 밀봉 밀수 밀착 밀폐 바느질 박멸 박제 반감 반납 반대 반려 반박 반복 반사 반영 반입 반전 반증 반추 반출 반포 반품 발간 발굴 발달 발동 발매 발산 발생 발아
발음 발의 발전 발주 발탁 발파 발포 발표 발현 방송 방수 방영 방임 방지 방출 방치 배가 배격 배부 배색 배설 배양 배출 배태 배포 배합 번복 번식 번역 벌채 변동 변명 변별 변색 변조 변혁 변형 변화 변환 병립 병치 병합 병행 보강 보관 보도
보류 보수 보완 보유 보전 보존 복개 복구 복권 복귀 복속 복용 복원 복제 복종 복직 복합 봉합 봉헌 부가 부각 부결 부담 부설 부식 부연 부인 부조 부착 부화 부활 부흥 분간 분담 분립 분별 분비 분사 분쇄 분실 분장 분절 분출 분화 불식 불입
불허 비약 비유 비준 비치 사료 사사 사수 사용 사유 사임 사장 삭감 산입 산정 산화 살균 살포 삽입 상기 상대 상비 상상 상설 상쇄 상술 상승 상신 상연 상영 상용 상장 상정 상징 색칠 생각 생략 생산 생성 생포 서술 석방 선결 선도 선발 선양
선언 선임 선적 선전 선정 선창 선포 선행 선회 설계 설립 설명 설복 설비 설정 설치 설파 섭취 성사 성숙 성안 성장 성찰 성취 세공 세분 세습 세척 세탁 소각 소급 소독 소망 소멸 소비 소실 소요 소유 소장 소진 소집 소탕 소화 속개 속칭 손실
- 105 -
손질 송부 송신 수감 수강 수거 수급 수긍 수렴 수록 수리 수립 수매 수상 수습 수장 수정 수집 수행 수확 숙성 순장 순화 스케치 스크랩 습득 승격 승계 승급 승진 시공 시달 시도 시동 시상 시술 시위 시작 시정 시찰 시청 시판 시행 식별 식용 신봉
신설 신축 실시 실용 실증 실추 실행 실현 심의 심화 안배 안장 안정 안치 암거래 압송 압축 앙양 애독 애용 애창 야기 약정 약칭 약화 양립 양분 양산 양성 양식 양육 양조 얘기 어림 어획 억제 언급 언명 엄금 엄단 엄벌 엄선 엄수 여과 역류 역설
역전 역행 연계 연구 연금 연마 연명 연상 연소 연습 연역 연임 연장 연재 연주 연출 열거 염려 염색 영속 영위 영입 영전 예감 예견 예금 예기 예매 예보 예비 예상 예속 예정 예증 예측 예치 예탁 오도 오발 오역 오용 오진 옹립 옹호 와전 와해
완결 완공 완료 완비 완수 완역 완치 완화 왜곡 요리 요망 요약 용납 용인 용접 용해 우려 우회 운구 운반 운송 운영 운용 운전 운항 운행 원용 위반 위배 위압 위장 위촉 유기 유도 유보 유예 유용 유인 유입 유지 유추 유출 유치 유통 유폐 유포
육성 융합 융해 은닉 은유 은폐 음미 응고 응용 응징 응축 의결 의도 의미 의식 의존 이동 이룩 이야기 이월 이입 이장 이첩 이체 이행 인상 인솔 인쇄 인용 인지 인하 인화 일관 일괄 일단락 일임 일축 임명 임신 입건 입금 입력 입법 입수 입안 입양
- 106 -
입학 잉태 자각 자제 자행 작곡 작성 작정 장려 장만 장사 장식 장악 장착 장치 재개 재건 재고 재단 재론 재배 재생 재연 재편 재현 쟁취 저금 저수 저술 저장 저축 적대 적립 적발 적시 적재 적중 전가 전개 전락 전래 전보 전사 전승 전역 전진
전치 전파 절감 절개 절상 절약 절제 절충 절판 절하 점거 점철 접맥 접목 접종 접지 접합 정간 정련 정렬 정선 정수 정의 정정 정제 정죄 정초 정화 제고 제기 제보 제본 제어 제창 제청 조각 조리 조립 조상 조성 조율 조음 조장 조절 조제 조준
조직 조차 조처 조치 조판 조합 조형 존립 존속 종결 종료 종식 종합 주도 주선 주장 주조 주지 주차 주창 주체 준공 준비 준수 중개 중건 중시 중용 중재 중첩 중화 증가 증강 증류 증발 증산 증설 증식 증액 증자 증진 증축 지각 지속 지양 지연
지체 지칭 지탱 지향 직감 직결 직선 직조 진보 진수 진술 진열 진작 진정 진행 진화 짐작 집결 집계 집약 집적 집중 집필 차감 차용 차입 차출 착각 착공 착색 착수 착안 착용 찬미 참작 참조 창간 창건 창달 창립 창설 창안 창업 창작 창제 창출
채굴 채록 채색 채용 채집 채취 채택 책봉 처리 척결 천거 천도 천명 천시 철수 철시 철폐 철회 첨가 첨부 첨삭 청구 청산 청소 체감 체결 체념 체득 체크 체현 체화 초극 초래 초빙 초연 초월 촉발 촉진 총괄 총칭 촬영 추가 추계 추구 추념 추대
- 107 -
추론 추산 추상 추서 추억 추적 추진 추출 추측 축성 축소 축약 축적 축조 출간 출고 출금 출동 출력 출발 출산 출원 출자 출제 출토 출품 출하 출현 충당 충원 충전 충족 취하 취합 측량 치장 치환 침식 침윤 침투 카피 캐스팅 컨트롤 코팅 타개 타산
타살 타이핑 타전 타파 타협 탄생 탈고 탈색 탈수 탈환 탐구 탐색 탐지 탑재 탕진 터득 토로 토벌 토의 통과 통용 통일 통찰 통칭 퇴거 퇴치 투고 투과 투매 투사 투시 투영 투척 투하 투합 특진 파병 파생 파손 파종 파직 판독 판명 판별 판서 패러디
편제 편찬 폄하 폐교 폐기 폐위 폐차 포박 포섭 포용 포획 폭락 폭로 표구 표기 표방 표백 표상 표시 표출 표현 풀이 풍자 프린트 피력 필사 하역 할증 합병 합산 합성 합창 항진 해결 해독 해동 해명 해부 해석 해설 해소 해지 행사 향상 향유 허비
헌상 헌정 혁파 현상 협의 호가 호도 호소 호송 호위 호전 호칭 호환 혼동 혼용 혼합 홍보 화장 화해 확대 확립 확보 확산 확장 확정 확충 환기 환산 환원 환전 활용 회복 회부 회상 회생 회전 회피 획득 후송 후회 훼방 흡입 흥분 흥행 희석
vn-tr-plv6
감사 감세 강습 격찬 결재 내사 냉대 다짐 당부 대우 독촉 동정 모함 박해 배상 변호 보조 부탁 분부 비방 사과 사랑 사찰 상환 선대 선망 선물 선불 수사 승낙 시주 압박 압제 언도 연수 원망 위안 인사 입찰 자백 장담 재가 재판 저주
- 108 -
조공 존경 종용 주문 증오 진찰 질책 천대 청부 청탁 추렴 축원 축하 충동 취조 치하 칭찬 타격 탕감 표창 핍박 하대 혐오 협박 협찬 호강 홀대 환대 환영 흠모 vn-tr-plv7
간파 갈취 강간 강점 강타 겁탈 격퇴 구타 기만 기습 난자 능욕 망신
면박 미행 배신 사기 사형 살상 선동 선점 성폭행 소박 수탈 신문 실점 양단 우롱 이유 이혼 저격 제압 즉사 징벌 차압 찬탈 참패 참형 추격 침입 커트 폭격 폭행 표류 혹사 화형 횡령
- 109 -
sign := feature-struct & [
ORTH *dlist*, MORPH morph,
SYN syn-cat, SEM sem-cat,
CTXT ctxt-cat, ARG-ST *list*, ARGS *list* ]. morph := feature-struct & [
PASS-TYPE pass-type ]. pass-type := feature-struct & [ SUFPASS bool, AUXPASS bool, PLVPASS bool ]. syn-cat := feature-struct & [ HEAD head, LEX bool, VAL val-cat ]. val-cat := feature-struct & [ SUBJ *list*, COMPS *list*, MOD *list* ].
Appendix II (Implementation)
- 110 -
head := feature-struct & [ POS pos, CASE case, FORM form, VERBAL bool, NOMINAL bool, PASSIVE bool, AUX bool, AGT bool, STATIVE bool, ANIMATE bool, ADVERSITY bool, TENSE tense ]. v-pass := v-alt & [
SYN.HEAD [ AGT -, PASSIVE + ], ARGS < v-tr & [MORPH.PASS-TYPE.SUFPASS + ] > ]. v-neutral-pass := v-pass & [
ARG-ST <
#2, #1 & [ SYN.HEAD.CASE.SCASE dat ] >, ARGS < [ ARG-ST < #1, #2 > ] > ]. inherent-passive-v := v-intr & [
MORPH.PASS-TYPE [ SUFPASS -, AUXPASS - ], SYN.HEAD.PASSIVE +, ARG-ST <
[ SYN.HEAD [ NOMINAL +, CASE [ GCASE nom, SCASE no_scase ] ] ], [ SYN.HEAD [ NOMINAL +, CASE.SCASE dat ] ]
> ].
- 111 -
hd-lex-ex := lex-ex & [ SYN.HEAD #1 ARGS < #2 & [ SYN.LEX + ], #1 & [ SYN [ HEAD.AUX +, VAL.COMPS #2 ] ] > ]. hd-lex-pass-ex := hd-lex-alt-ex & [ SYN.VAL [ SUBJ #2, COMPS #1 ], ARGS < [ MORPH.PASS-TYPE.AUXPASS +, ARG-ST <#1, #2> ], [ SYN [ HEAD.PASSIVE +, VAL.SUBJ #2 ] > ].
- 112 -
vn-passive-light-v := vn-light-v & [
SYN.HEAD.PASSIVE +, SEM #sem, ARG-ST
< #2, [ SYN.HEAD.CASE.SCASE dat ] >, ARGS < vn-tr & [ MORPH.PASS-TYPE.PLVPASS +, ARG-ST < [ ], #2 >, SEM #sem ] > ]. vn-passive-light-v-toy := vn-passive-light-v & [ SYN.HEAD.AGT -, ARGS < [ SYN.HEAD.ANIMATE - ] > ]. vn-passive-light-v-pat := vn-passive-light-v & [ SYN.HEAD.AGT +, ARGS < [ SYN.HEAD [ ANIMATE +, ADVERSITY - ] ] > ]. vn-passive-light-v-tangha := vn-passive-light-v & [ SYN.HEAD.AGT +, ARGS < [ SYN.HEAD [ ANIMATE +, ADVERSITY + ] ] > ].
- 113 -
pass-light-v := light-v & [
SYN.HEAD.PASSIVE + ]. pass-light-toy := pass-light-v & [
SYN.HEAD [ AGT -, ANIMATE - ], ARG-ST.REST.FIRST.SYN.HEAD.CASE [ GCASE nom, SCASE no_scase ]
]. pass-light-ani := pass-light-v & [
SYN.HEAD [ AGT +, ANIMATE + ], ARG-ST.REST.FIRST.SYN.HEAD.CASE [ GCASE vcase, SCAS E no_scase ]
]. pass-light-pat := pass-light-ani & [
SYN.HEAD.ADVERSITY – ]. pass-light-tangha := pass-light-ani & [
SYN.HEAD.ADVERSITY + ].
- 114 -
hd-lex-vn-pass-ex := hd-lex-alt-ex & [ SYN [ HEAD.PASSIVE +, VAL [ SUBJ < #2 & [ SYN.HEAD [ CASE [ GCASE nom, SCA SE no_scase ] ] ] >, COMPS.FIRST #1 & [ SYN.HEAD.CASE.SCASE dat ] ] ], ARGS < [ MORPH.PASS-TYPE.PLVPASS +, SYN [ HEAD [ ANIMATE #animate, ADVERSITY #adversit y ], VAL [ SUBJ #1, COMPS.FIRST #2 & [ SYN.HEAD.C ASE.GCASE acc] ] ] ], [
SYN.HEAD [
VERBAL +, PASSIVE +, ANIMATE #animate, ADVERSITY #adversity
] ] ]
> ].