dialogue based meaning negotiation
TRANSCRIPT
Dialogue Based Meaning Negotiation
Gabrielle Santos, Valentina Tamma,Terry R. Payne & Floriana Grasso
University of Liverpool
[email protected]@liverpool.ac.uk
[email protected]@liverpool.ac.uk
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alignment
Correspondence
Open Systems, Ontologies and Alignment
• Agents can assume different ontological models• Modelled implicitly, or explicitly by defining entities (classes, roles etc),
typically using some logical theory, i.e. an Ontology
• Alignment Systems align similar ontologies
2
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Differences in Modelling• Different ways of modelling ontologies can generate
difficulties when trying to make an alignment.
3
Cow
Ontology 1
Farm Animal
Cow
Ontology 2
Herbivore
Soccer Football• Terminological heterogeneity:
• Different terms represent the same entity
• Structural heterogeneity:• The same concept is modelled differently from
one ontology to another. • Ontology 1: A Cow is a Farm Animal.• Ontology 2: A Cow is a Herbivore.
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
RDF/OWL Ontologies
• Concepts and properties are modelled as a graph• Concepts modelled as nodes
• Relationships between concepts are edges
• Triples represent a concept that has a relationship to another one • subject - predicate - object
4
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
RDF/OWL Ontologies
• Concepts and properties are modelled as a graph• Concepts modelled as nodes
• Relationships between concepts are edges
• Triples represent a concept that has a relationship to another one • subject - predicate - object
5
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
RDF/OWL Ontologies
• Concepts and properties are modelled as a graph• Concepts modelled as nodes
• Relationships between concepts are edges
• Triples represent a concept that has a relationship to another one • subject - predicate - object
6
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Align Everything?
• Do we need everything to be aligned?• An agent may aggregate several ontologies for a variety of domains
• A task may be relevant to only a single fragment within an ontology
• Fragments of the ontological space may be confidential, or commercially sensitive.
• We focus only on a specific signature (set of ontological entities) relevant to a transaction.
7
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Aligning with no prior knowledge ?
• Can two agents exchange partial knowledge to find a meaningful alignment?
• Existing approaches assume: • Full disclosure of knowledge to a 3rd party • Existence of existing alignment fragments • Knowledge of underlying grounds for correspondence generation
• What if there is no prior knowledge?8
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Meaning-based Dialogue
• Dialogue fragment that…• Allows two agents to exchange knowledge about entities and
their relationships to agree upon a mutually acceptable final alignment.
• Disclose only those entities (T-Box) directly related to the correspondence
• Minimises the disclosure of redundant or irrelevant entities.9
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Meaning-based Dialogue
• Explores a cognitive approach to reaching consensus over possible correspondences…• Agents identify possible concepts that may be
ontologically equivalent in their respective ontologies • Each then seeks further evidence over the locality of each
concept to verify if these are structurally similar. • Both agents have the opportunity to ask questions • Correspondences only accepted if both agents accept the
same underlying support
10
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Meaning-based Dialogue• Inherently heuristic
• Need to balance exploration vs exploitation • Aim to find a pragmatic, if sub-optimal, solution that is fit for
purpose
• If all knowledge is disclosed, then could use 3rd party approach to find optimal solution
• If insufficient knowledge is disclosed, this could increase occurrence of false positives
• Depth of exploration is significant • Shallow structure similarity can fail due to modelling differences • However, deeper structural similarities hay have less relevance
11
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Roadmap• Work to date has developed
• Cognitive framework that underpins the dialogue
• Dialogue fragment targeted at allowing two agents to agree on a single correspondence
• Protocol / Finite State Machine describing the permissible moves during the dialogue
• Practical reasoning approach to determine what moves to make at each state - this however is dependent on the objective function
• Work presented here covers• Underlying cognitive model
• State machine and moves
12
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Cognitive Approach
13
Open
Propose
Close
Confirm
Agent initiates the dialogue • Identify the name of the entity to be aligned. • Interlocutor can chose to offer a candidate
corresponding entity, or reject the dialogue
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Propose
14
Open
Propose
Close
Confirm
Allows agent to gather supporting evidence in favour of a correspondence
• Asks for triples representing the entities locality • Attempts to match this to its own triples
If sufficient support found, then it can assert a correspondence
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Confirm
15
Open
Propose
Close
Confirm
Interlocutor verifies the support for the correspondence
• Can accept the correspondence • May ascertain its own evidence to augment the
support (i.e returning to the propose phase)
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Close
16
Open
Propose
Close
ConfirmAlice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue fragment terminates if • Both agents accept the correspondence • No support can be found that is acceptable to
both agents
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Roadmap (cont)
• We assume (for now)…1. Candidate concepts will be lexically similar
• Lexical Similarity Metrics exist - black box approach
2. Structural similarity can be found on partial ontological fragments • Such metrics exist (such as similarity flooding), but unclear how to
determine depth of exploration
3. Agents will have different knowledge bases, and possibly different similarity metrics • What holds for one agent, cannot be guaranteed to hold for the other
17
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Roadmap (cont)• We assume (for now)…
4. Concept relationships (triples) will be disclosed in order of significance • Mechanism to establish this is subject of other work
5. Threshold is a function of the size of the support, and non-linear • If we assume that the most significant triple is
disclosed, but is not sufficiently convincing, then subsequent evidence may strengthen the support, but not increase the quantitative mean
• A function has been selected that reduces the threshold asymptotically as the number of triple pairs (x) increases
18
"(x) = 12(x+1) + 0.5
0.5
0.6
0.7
0.8
0.9
1
0 5 10 15 20
f(x)
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue Moves: initiate & fail
• These moves are used to initiate or close (through failure) the dialogue• <a, initiate, e, nil, nil>
• Agent a starts the dialogue fragment by offering the name of the source entity e
• <a, fail, e, nil, nil> • Agent a cannot find a candidate destination entity to map to the source
entity e
19
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue Moves: propose & rejectC
• These moves are used to propose or reject a possible correspondence between entities e and e’• <a, propose, e, e’, nil>
• Agent a suggests a candidate destination entity e’ as one that matches the source entity e. However, no support yet exists to assert this as a correspondence.
• <a, rejectC, e, e’, nil> • Agent a has failed to find any convincing evidence to accept a
candidate correspondence between e and e’.
20
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue Moves: justify & testify
• These moves are used to propose or reject a possible correspondence between entities e and e’• <a, justify, e, e’, nil>
• Agent a requests a new triple to support the candidate correspondence.
• <a, testify, e, e’, 𝜛> • Agent a responds to a justify request by providing a (previously
undisclosed) triple 𝜛 based on the locality of the entity (e or e’) in its ontology.
21
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue Moves: assert & rejectS
• These moves are used to assert or reject a possible correspondence based on supporting evidence• <a, assert, e, e’, 𝛼>
• Agent a asserts the argument 𝛼 for a candidate correspondence given a subset of triples believed to be matching (from agent a’s perspective).
• <a, rejectS, e, e’, 𝛼> • Agent a rejects the argument 𝛼 as the support was not believed
convincing (from that agent’s perspective).
22
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
An argument 𝝰 is a tuple (S, c) where:S = set of matching pairs of triples c = candidate correspondence
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue Moves: accept
• This move confirms the acceptability of the correspondence• <a, accept, e, e’, 𝛼>
• Agent a accepts the support, and hence argument 𝛼 for a candidate correspondence between e and e’.
23
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Alice3A
Bob2B
Alice1A
Bob4B
Alice6A
Alice8A
Bob5B
Bob7B
fail
initiate
rejectC
propose
rejectC
justify
testify
justify
justify
assert
rejectS
justify
testify
assertrejectS
accept
accept
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Example Dialogue
24
Alice’s PrivateKnowledge
As the support is below threshold (from Alice’s perspective), she requests further evidence to support the candidate correspondence between d and w.
Commitment Store CS (Public Knowledge)
Bob’s Private KnowledgeOntology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Alice’s $ Bob’s $ �Alices
hd, l, gi hw, t, yi 0.66
Gamma Store Γ Gamma Store Γ
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
S = {(hd, l, gi, hw, t, yi)}�̄Alices (S) = 0.66
"(|S|) = 0.75 ✘
"(x) = 12(x+1) + 0.5
0.5
0.6
0.7
0.8
0.9
1
0 5 10 15 20
f(x)
Dialogue based Meaning NegotiationTerry Payne University of Liverpool 25
Commitment Store CS (Public Knowledge)Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Bob responds with another triple in the locality of the concept w. Alice now has sufficient support to assert the correspondence.
Gamma Store Γ
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Alice’s $ Bob’s $ �Alices
hd, l, gi hw, t, yi 0.66hd, k, ei hw, r, zi 0.70
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Alices (S) = 0.68
"(|S|) = 0.6̇ ✔
Alice’s PrivateKnowledge
Gamma Store Γ
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Bob’s Private Knowledge
Dialogue based Meaning NegotiationTerry Payne University of Liverpool 26
Commitment Store CS (Public Knowledge)Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Alice makes the assertion; however, from Bob’s perspective, the support is below threshold.
Gamma Store ΓGamma Store Γ
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Alice’s $ Bob’s $ �Alices
hd, l, gi hw, t, yi 0.66hd, k, ei hw, r, zi 0.70
Alice’s $ Bob’s $ �Bob
s
hd, l, gi hw, t, yi 0.60hd, k, ei hw, r, zi 0.68
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Alices (S) = 0.68
"(|S|) = 0.6̇ ✔ ✘
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Bob
s (S) = 0.64"(|S|) = 0.6̇
Alice’s PrivateKnowledge
Bob’s Private Knowledge
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Gamma Store ΓGamma Store Γ
27
Commitment Store CS (Public Knowledge)Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Alice’s $ Bob’s $ �Alices
hd, l, gi hw, t, yi 0.66hd, k, ei hw, r, zi 0.70
Alice’s $ Bob’s $ �Bob
s
hd, l, gi hw, t, yi 0.60hd, k, ei hw, r, zi 0.68
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Alices (S) = 0.68
"(|S|) = 0.6̇ ✔ ✘
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Bob
s (S) = 0.64"(|S|) = 0.6̇
It is now Bob’s turn to request additional evidence from Alice to support the candidate correspondence between d and w.
Alice’s PrivateKnowledge
Bob’s Private Knowledge
Dialogue based Meaning NegotiationTerry Payne University of Liverpool 28
Commitment Store CS (Public Knowledge)Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Move Locution
1 hAlice, initiate, d,nil ,nili2 hBob, propose, d, w,nili3 hAlice, justify , d, w,nili4 hBob, testify , d, w, hw, t, yii5 hAlice, justify , d, w,nili6 hBob, testify , d, w, hw, r, zii7 hAlice, assert, d, w, ({(hd, l, gi, hw, t, yi), (hd, k, ei, hw, r, zi)}, hd,w,⌘i)i8 hBob, justify , d, w,nili9 hAlice, testify , d, w, hd,m, fii10 hBob, assert, d, w, ({(hw, t, yi, hd, l, gi), (hw, r, zi, hd, k, ei), (hw, s, xi, hd,m, fi)},
hd,w,⌘i)i11 hAlice, accept, d, w, hd,w,⌘i)
Alice’s PrivateKnowledge
Bob’s Private Knowledge
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Ontology
d
e fw
y
z
t
Alice's
Ontology Bob's
Ontology
k
l
m
xg
s
r
n h
Gamma Store Γ Gamma Store Γ
✔
Alice responds with another triple in the locality of the concept d. Both Bob and Alice have sufficient support to assert and accept the correspondence.
Alice’s $ Bob’s $ �
Bob
s
hd, l, gi hw, t, yi 0.60hd, k, ei hw, r, zi 0.68hd,m, fi hw, s, xi 0.61
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi),(hd,m, fi, hw, s, xi)}
�̄
Bob
s (S) = 0.63"(|S|) = 0.625 ✔
S = {(hd, l, gi, hw, t, yi),(hd, k, ei, hw, r, zi)}
�̄Alices (S) = 0.68
"(|S|) = 0.6̇
Alice’s $ Bob’s $ �Alices
hd, l, gi hw, t, yi 0.66hd, k, ei hw, r, zi 0.70
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Ongoing work• Formally defining / validating the dialogue
• Currently only the performatives / moves have been defined. • The decision-making process at each step of the protocol is being defined
for possible objective functions • Minimising disclosure • Minimising support etc
• Determining similarity metrics• Several approaches to determining:
• lexical similarity - simple !!! • structural similarity - not so easy…
• Ranking triples for disclosure is more tricky (objective-function dependent)
29
Dialogue based Meaning NegotiationTerry Payne University of Liverpool
Ongoing work (cont)• Expanding the dialogue beyond
single correspondences• Current fragment handles exchange of
triples pertaining to the locality of one candidate correspondence (i.e. singleton signature).
• Need to extend this for signatures for multiple entities.
• Include ability to accept existing, established correspondences.
• Resolve ambiguous alignments (re-negotiate if multiple candidates are possible)
30
publication article author
submittedPaper reviewedPaper paper editor
A Dialectical Approach to Selectively Reusing Ontological CorrespondencesValentina Tamma University of Liverpool
Questions
31
For other papers on this and our other related work:
http://www.csc.liv.ac.uk/~trp/Knowledge-Based-Agents.html