gate, human language and machine learning gate.ac.uk/ nlp.shef.ac.uk

30
GATE, Human Language and Machine Learning http://gate.ac.uk/ http://nlp.shef.ac.uk/ Hamish Cunningham, Valentin Tablan, Kalina Bontcheva, Diana Maynard 9 th July/2003 1. The Knowledge Economy and Human Language Technology 2. GATE: a General Architecture for Text Engineering 3. GATE, Information Extraction and Machine Learning

Upload: etana

Post on 09-Jan-2016

55 views

Category:

Documents


3 download

DESCRIPTION

GATE, Human Language and Machine Learning http://gate.ac.uk/ http://nlp.shef.ac.uk/ Hamish Cunningham, Valentin Tablan, Kalina Bontcheva, Diana Maynard 9 th July/2003 The Knowledge Economy and Human Language Technology GATE: a General Architecture for Text Engineering - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

GATE, Human Language and Machine Learning

http://gate.ac.uk/ http://nlp.shef.ac.uk/

Hamish Cunningham, Valentin Tablan, Kalina Bontcheva, Diana Maynard

9th July/2003

1. The Knowledge Economy and Human Language Technology2. GATE: a General Architecture for Text Engineering 3. GATE, Information Extraction and Machine Learning

Page 2: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

2(30)

                                                                                                                           

1. The Knowledge Economy and Human Language

Gartner, December 2002:

• taxonomic and hierachical knowledge mapping and indexing will be prevalent in almost all information-rich applications

• through 2012 more than 95% of human-to-computer information input will involve textual language

A contradiction: formal knowledge in semantics-based systems vs. ambiguous informal natural language

The challenge: to reconcile these two opposing tendencies

Page 3: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

3(30)

Information Extraction (1): from text to structured data

Two trends in the early 1990s:• NLU: too difficult! Restrict the task and increase the performance• Quantitative measurement (MUC – Message Understanding Conference, ACE –

Advanced Content Extraction, TREC – Text Retrieval Conference...) means good estimation of accuracy

Types of extraction:• Identify named entities (domain independent)

• Persons• Dates• Numbers• Organizations

• Identify domain-specific events and terms; e.g., if we’re processing football:• Relations: which team a player plays for• Events: goal, foul, etc

Page 4: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

4(30)

Information Extraction (2)

MUC-7 tasks• NE: Named Entity recognition and typing • CO: co-reference resolution • TE: Template Elements • TR: Template Relations • ST: Scenario Templates

Example: The shiny red rocket was fired on Tuesday. It is the brainchild of Dr. Big Head. Dr. Head is a staff

scientist at We Build Rockets Inc.

• NE: entities are "rocket", "Tuesday", "Dr. Head" and "We Build Rockets" • CO: "it" refers to the rocket; "Dr. Head" and "Dr. Big Head“ are the same • TE: the rocket is "shiny red" and Head's "brainchild". • TR: Dr. Head works for We Build Rockets Inc. • ST: a rocket launching event occurred with the various participants.

Page 5: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

5(30)

HumanLanguage

Formal Knowledge(ontologies andinstance bases)

(A)IE

CLIE

(M)NLG

ControlledLanguage

OIE

SemanticWeb; Semantic Grid;Semantic Web Services

KEYMNLG: Multilingual Natural Language GenerationOIE: Ontology-aware Information ExtractionAIE: Adaptive IECLIE: Controlled Language IE

IE and Knowledge: Closing theLanguageLoop

Page 6: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

6(30)

Populating Ontologies with IE

Page 7: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

7(30)

Protégé and Ontology Management

Page 8: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

8(30)

IE: the bad news…

Domain specificity vs. task complexity:

complexity

spec

ific

ity “acceptable” accuracy

domainspecific

simple entities

events and relations

verygeneral

Page 9: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

9(30)

                                                                                                                           

2. GATE: Software Architecure for HLT

Software lifecycle in collaborative research

Project Proposal: We love each other. We can work so well together. We can hold workshops on Santorini together. We will solve all the problems of AI that our predecessors were too stupid to.

Analysis and Design: Stop work entirely, for a period of reflection and recuperation following the stress of attending the kick-off meeting in Luxembourg.

Implementation: Each developer partner tries to convince the others that program X that they just happen to have lying around on a dusty disk-drive meets the project objectives exactly and should form the centrepiece of the demonstrator.

Integration and Testing: The lead partner gets desperate and decides to hard-code the results for a small set of examples into the demonstrator, and have a fail-safe crash facility for unknown input ("well, you know, it's still a prototype...").

Evaluation: Everyone says how nice it is, how it solves all sorts of terribly hard problems, and how if we had another grant we could go on to transform information processing the World over (or at least the European business travel industry).

Page 10: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

10(30)

                                                                                                                           

GATE, a General Architecture for Text Engineering• An architectureA macro-level organisational picture for LE software systems. • A frameworkFor programmers, GATE is an object-oriented class library that implements the architecture. • A development environmentFor language engineers, computational linguists et al, GATE is a graphical development environment bundled with a set of tools for doing e.g. Information Extraction. • Some free components... ...and wrappers for other people's components • Tools for: evaluation; visualise/edit; persistence; IR; IE; dialogue; ontologies; etc.• Free software (LGPL). Download at http://gate.ac.uk/download/

Page 11: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

11(30)

                                                                                                                           

Architectural principles

• Non-prescriptive, theory neutral (strength and weakness) • Re-use, interoperation, not reimplementation (e.g. diverse XML support, integration of tools like Protégé, Jena and Weka) • (Almost) everything is a component, and component sets are user-extendable

Component-based development

• An OO way of chunking software: Java Beans • GATE components: CREOLE = modified Java Beans (Collection of REusable Objects for Language Engineering) • The minimal component = 10 lines of Java, 10 lines of XML, 1 URL.

Page 12: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

12(30)

                                                                                                                           

GATE Language ResourcesGATE LRs are documents, ontologies, corpora, lexicons, ……

Documents / corpora:• GATE documents loaded from local files or the web... • Diverse document formats: text, html, XML, email, RTF, SGML.

Processing ResourcesAlgorithmic components knows as PRs – beans with execute methods.• All PRs can handle Unicode data by default. • Clear distinction between code and data (simple repurposing).• 20-30 freebies with GATE• e.g. Named entity recognition; WordNet; Protégé; Ontology; OntoGazetteer; DAML+OIL export; Information Retrieval based on Lucene

Page 13: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

13(30)

Vis

ual

Res

ourc

es

Page 14: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

14(30)

 Performance Evaluation

• At document level – annotation diff

• At corpus level – corpus benchmark tool – tracking system’s performance over time

Page 15: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

15(30)

Regression Test – Corpus Benchmark Tool

Page 16: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

16(30)

Information Retrieval

Based on the Lucene IR engine

Page 17: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

17(30)

                     

GATE Unicode Kit (GUK) Java provides no special support for text input (this may change)

• Support for defining additional Input Methods (IMs)

• currently 30 IMs for 17 languages

• Pluggable in other applications

Editing Multilingual Data

Page 18: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

18(30)

Processing Multilingual Data

All the visualisation and editing tools for ML LRs use enhanced Java facilities:

Page 19: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

19(30)

A bit of a nuisance (users)GATE team projects:• Conceptual indexing: MUMIS:

automatic semantic indices for sports video

• MUSE, cross-genre entitiy finder• HSL, Health-and-safety IE• ETCSL: collaboration with IOAS

Oxford on Sumerian• Old Bailey: collaboration with HRI

on 17th century court reports• Multiflora: plant taxonomy text

analysis for biodiversity research e-science

• Advanced Knowledge Technologies: €12m UK five site collaborative project

• H-TechSight: knowledge portal for Chemicals Engineers

• Framework 6: SEKT, PrestoSpace, KnowledgeWeb

A representative fraction of GATE users: • IBM TJ Watson, US• the American National Corpus

project, US • the Perseus Digital Library project,

Tufts University, US• Longman Pearson publishing, UK• Merck KgAa, Germany• Canon Europe, UK• Knight Ridder (the second biggest

US news publisher)• BBN (leading HLT research lab), US• SMEs in Sirma AI Ltd., Bulgaria• Imperial College, London, the

University of Manchester, the University of Karlsruhe, Vassar College, the University of Southern California and a large number of other UK, US and EU Universities

• UK and EU projects inc. MyGrid, CLEF, dotkom, AMITIES, Cub Reporter, EMILLE, MUSE, Poesia...

Page 20: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

20(30)

3. Machine Learning in GATE

• Uses classification.[Attr1, Attr2, Attr3, … Attrn] Class

• Classifies annotations.(Documents can be classified as well using a simple

trick.)

• Annotations of a particular type are selected as instances.

• Attributes refer to instance annotations.• Attributes have a position relative to the instance

annotation they refer to.

Page 21: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

21(30)

Attributes

Attributes can be:– Boolean

The [lack of] presence of an annotation of a particular type [partially] overlapping the referred instance annotation.

– NominalThe value of a particular feature of the referred instance annotation.

The complete set of acceptable values must be specified a-priori.

– NumericThe numeric value (converted from String) of a particular feature of

the referred instance annotation.

Page 22: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

22(30)

Implementation

Machine Learning PR in GATE.Has two functioning modes:

– training– application

Uses an XML file for configuration:<?xml version="1.0" encoding="windows-1252"?>

<ML-CONFIG><DATASET> … </DATASET><ENGINE>…</ENGINE>

<ML-CONFIG>

Page 23: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

23(30)

<DATASET>

<DATASET><INSTANCE-TYPE>Token</INSTANCE-TYPE> <ATTRIBUTE> <NAME>POS_category(0)</NAME> <TYPE>Token</TYPE> <FEATURE>category</FEATURE> <POSITION>0</POSITION> <VALUES> <VALUE>NN</VALUE> <VALUE>NNP</VALUE> <VALUE>NNPS</VALUE> … </VALUES> [<CLASS/>] </ATTRIBUTE> …</DATASET>

Page 24: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

24(30)

<ENGINE>

<ENGINE> <WRAPPER>gate.creole.ml.weka.Wrapper</WRAPPER> <OPTIONS> <CLASSIFIER>weka.classifiers.j48.J48</CLASSIFIER> <CLASSIFIER-OPTIONS>-K 3</CLASSIFIER-OPTIONS> <CONFIDENCE-THRESHOLD>0.85</CONFIDENCE-THRESHOLD> </OPTIONS> </ENGINE>

Page 25: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

25(30)

Attributes Position

Instances type: Token

Page 26: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

26(30)

Machine Learning PR

• Can save a learnt model to an external file for later use.Saves the actual model and the collected dataset.

• Can export the collected dataset in .arff format.

Page 27: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

27(30)

Standard Use Scenario

Training• Prepare training data by enriching

the documents with annotation for attributes. (e.g. run Tokeniser, POS tagger, Gazetteer, etc).

• Run the ML PR in training mode.• Export the dataset as .arff and

perform experiments using the WEKA interface in order to find the best attribute set / algorithm / algorithm options.

• Update the configuration file accordingly.

• Run the ML PR again to collect the actual data.

• [ Save the learnt model. ]

Application• Prepare data by enriching the

documents with annotation for attributes. (e.g. run Tokeniser, POS tagger, Gazetteer, etc).

• [ Load the previously saved model. ]

• Run the ML PR in application mode.

• [ Save the learnt model. ]

Page 28: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

28(30)

An Example

Learn POS category from POS context.

Page 29: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

29(30)

Using Other ML Libraries

The MLEngine InterfaceMethod Summary• void addTrainingInstance(List attributes)

Adds a new training instance to the dataset. • Object classifyInstance(List attributes)

Classifies a new instance. • void init()

This method will be called after an engine is created and has its dataset and options set. 

• void setDatasetDefinition(DatasetDefintion definition) Sets the definition for the dataset used. 

• void setOptions(org.jdom.Element options) Sets the options from an XML JDom element.

• void setOwnerPR(ProcessingResource pr) Registers the PR using the engine with the engine. 

Page 30: GATE, Human Language and  Machine Learning gate.ac.uk/ nlp.shef.ac.uk

30(30)

                                                                                                                           

Conclusion

GATE is:• Addressing the need for scalable, reusable, and portable HLT solutions• Supporting large data, in multiple media, languages, formats, and locations• Lowering the cost of creation of new language processing components • Promoting quantitative evaluation metrics via tools and a level playing field• Promoting experimental repeatability by developing and supporting free

software

Perhaps it may become:• A vehicle for the spread of collaborative experiments in ML and HLT?