Download presentation
Presentation is loading. Please wait.
Published byRegina Powers Modified over 9 years ago
1
1 Special Electives of Comp.Linguistics: Processing Anaphoric Expressions Eleni Miltsakaki AUTH Fall 2005-Lecture 7
2
2 Plan for today Coherence, Reference and the Theory of Grammar Andrew Kehler (2000)
3
3 A theory of discourse coherence Background –Hobbs (1990) –Literature and Cognition –Hume (1748) – An Inquiry Concerning Human Understanding Coherence relations –Resemblance –Cause-effect –Contiguity
4
4 Resemblance Parallel Contrast Exemplification Generalization Exception Elaboration
5
5 Parallel Infer p(a1, a2, …) from the assertion of S1 and p(b1, b2, …) from the assertion of S2, where for some property vector q, qi(ai) and qi(bi) for all i 1.Dick Gephardt organized rallies for Gore, and Tom Daschle distributed pamphlets for him P1= organized rallies for P2= distributed pamphlets P= do something to support A1= Dick B1= Tom
6
6 Contrast Infer p(a1,a2, …) from the assertion of S1 and ~p(b1, b2, …) from the assertion S2, in which for some property vector q, qi(ai) and qi(bi) for all i 1.Gephardt supported Gore, but Armey opposed him
7
7 Cause-Effect Result Explanation Violated expectation Denial of preventer
8
8 Result Infer P from the assertion of S1 and Q from the assertion of S2, where normally P Q 1.George is a politician, and therefore he’s dishonest
9
9 Explanation Infer P from the assertion of S1 and Q from the assertion of S2, where normally Q P 1.George is dishonest because he’s a politician 2.George is dishonest. He’s a politician
10
10 Violated expectation Infer P from the assertion of S1 and Q from the assertion of S2, where normally P ~Q 1.George is a politician, but he’s honest
11
11 Denial of preventer Infer P from the assertion of S1 and Q from the assertion of S2, where normally Q ~P 1.George is honest, even though he’s a politician
12
12 Contiguity Occasion (1) –Infer a change of state for a system of entities from S1, inferring the final state for from S2 Occasion (2) –Infer a change of state for a system of entities from S2, inferring the initial state for this system from S1 1.George picked up the speech. He began to read 2.Larry went into a restaurant. The baked salmon sounded good and he ordered it.
13
13 Coherence and pronoun interpretation Data to account for: 1.John kicked Bill. Mary told him to go home. [him=John] 2.Bill was kicked by John. Mary told him to go home. [him=Bill] 3.John kicked Bill. Mary punched him. [him=Bill]
14
14 A coherence driven approach (Hobbs 1979) 1.The city council denied the demonstrators a permit because… 1.… they feared violence 2.… they advocated violence (Winograd 1972) Explanation: Infer P from the assertion of S1 and Q from the assertion S2, where normally Q P World knowledge: Fear(X,V) & advocate(Y,V) & enable_to_cause(Z,Y,V) IMPLIES deny(X,Y,Z) Resolution Deny(city_council, demonstratros, permit) -- 1 Fear(T, violence) -- T=X=city council -- 1.1 Advocate (T, violence) – T=Y=demonstrators –1.2
15
15 Attention driven theories Centering theory –(captures the fact the speaker’s choice of referring expression affects the coherence of a discourse – this is not possible in a purely coherence-driven theory)
16
16 Parallelism Carl is talking to Tom in the lab. Terry wants to talk to him too. (parallelism cannot be captured in Centering)
17
17 Pattern in the data Data in support of attention-driven theories are related to the CONTIGUITY relation
18
18 Data in support of parallelism are related to the RESEMBLANCE relation
19
19 Data in support of coherence-driven theories are related to CAUSE-EFFECT relations
20
20 “The three types of data used to provide evidence for three corresponding approaches to pronoun resolution appear to be mutually contradictory when viewed from the purview of current theories ut in fact pattern with my neoHumian categorization of coherence relations” p.154
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.