where to next? pronoun interpretation as a side effect of discourse direction

25
Where to next? Pronoun interpretation as a side effect of discourse direction Hannah Rohde, Andy Kehler, & Jeff Elman UC San Diego UNY Conference on Human Sentence Processing, March 29-31 2007

Upload: indra

Post on 11-Feb-2016

15 views

Category:

Documents


0 download

DESCRIPTION

Where to next? Pronoun interpretation as a side effect of discourse direction. Hannah Rohde, Andy Kehler, & Jeff Elman UC San Diego. CUNY Conference on Human Sentence Processing, March 29-31 2007. Bill. He. John. He. Transfer Verb. Goal (to-phrase). Ambiguous Pronoun Prompt. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Where to next?    Pronoun interpretation as a   side effect of discourse direction

Where to next? Pronoun interpretation as a side effect of discourse direction

Hannah Rohde, Andy Kehler, & Jeff ElmanUC San Diego

CUNY Conference on Human Sentence Processing, March 29-31 2007

Page 2: Where to next?    Pronoun interpretation as a   side effect of discourse direction

2

Transfer of possession (Stevenson et al. 1994)

(1) handed a book to . ______________.John Bill HeJohn He recommended it

Transfer Verb

Source (subject)

Goal (to-phrase)

AmbiguousPronoun Prompt

thanked JohnBill He

50/50: Goal continuations / Source continuations No subject preference or grammatical parallelism Two explanations considered:

Thematic Role Preference Event Structure Bias

Page 3: Where to next?    Pronoun interpretation as a   side effect of discourse direction

3

Outline

Background: Rohde et al. 2006 Test Thematic-Role and Event-Structure biases Alternative account: Discourse Coherence

Experiments 1 & 2: test predictions of a coherence- based model using story continuations

Preliminary results: discourse effects in relative clause attachment

Page 4: Where to next?    Pronoun interpretation as a   side effect of discourse direction

4

(2) handed a book to . ______ .JohnSOURCE BillGOAL He(3) was handing a book to . ___ .JohnSOURCE BillGOAL He

Explaining salience of Goal (Rohde et al. 2006)

0%

20%

40%

60%

80%

100%

Perfective Imperfective

% Interpretation

SourceGoal

Goal bias ~ side effect of Event Structure

Thematic role preference or event structure bias?

Equivalent thematic roles but different event structure

Effect of aspectF(1,48)=50.622p<0.0001

Page 5: Where to next?    Pronoun interpretation as a   side effect of discourse direction

5

Establishing coherence: infer a relationship between the meanings expressed by two sentences (P&Q below) (Hobbs 1979, Kehler 2002)

Effects of coherence (Rohde et al. 2006)

(5) Matt passed a sandwich to David. He said thanks.

[Result: P Q]

David He

(4) Matt passed a sandwich to David. He didn’t want David to starve.

[Explanation: Q P]

Matt He

Causal relations (Explanation, Result, Violated Expectation)

Page 6: Where to next?    Pronoun interpretation as a   side effect of discourse direction

6

(6) Matt passed a sandwich to David. He did so carefully.

[Elaboration: infer P from both S1 and S2]

Coherence cont.

[Occasion: infer initial state of event described in S2 to be final state of event described in S1]

(7) Matt passed a sandwich to David. He ate it up.David He

Matt He

Similarity relations (Parallel, Elaboration)

Contiguity relations (Occasion)

Page 7: Where to next?    Pronoun interpretation as a   side effect of discourse direction

7

Goal bias following perfective context sentences limited to Occasion & Result (see Arnold 2001)

Interpretation as side effect of coherence distribution

Perfective Context Sentences

050

100150200

OccasionElaborationExplanationViolated-ExpResultParallel

CountSourceGoal

Discourse coherence effects (Rohde et al.)

Page 8: Where to next?    Pronoun interpretation as a   side effect of discourse direction

8

Test predictions of a coherence-driven model More Occasion/Result more Goal resolutions More Explanation/Elaboration/Violated-Exp more

Source

Shift coherence shift interpretation

Page 9: Where to next?    Pronoun interpretation as a   side effect of discourse direction

9

Experiment 1: objects-of-transfer

(8) John handed a book to Bill. He ___________ .

(9) John handed a bloody meat cleaver to Bill. He __ .

Proposal: elicit different continuations with different objects Stimuli: normal and bizarre objects

Predictions: If… Abnormal objects more Explanations and Explanations Source bias More Source continuations for (9) than (8)

Page 10: Where to next?    Pronoun interpretation as a   side effect of discourse direction

10

Methodology

Subjects: 69 monolingual English speakers Task: write 50 continuations, just like Rohde et al. Stimuli: 21 transfer-of-possession like Rohde et al.

(+ bizarre objects) Evaluation: judges assess coherence/interpretation Analysis:

Effect of within-subject factor of Object Type on Coherence (Elab/Expl/Occ/Par/Res/Viol-Exp) Pronoun interpretation (Source/Goal)

Mixed-effects logistic regression Controls for random effects of Subject and Item

Page 11: Where to next?    Pronoun interpretation as a   side effect of discourse direction

11

0%

20%

40%

60%

80%

100%

Normal Abnormal

% Coherence

ExpElabV-EResOccPar

Coherence varies by object

Results

0.81

0.820.99

0.050.17

0.71

Consistent prob(Source|coh)Exp 1Rohde et al.Coherence

Elaboration

0.87Violated-expectation

0.75Explanation0.99

0.16Result0.20Occasion

0.45Parallel

p<0.0001Source Source

Goal} Goal}

Page 12: Where to next?    Pronoun interpretation as a   side effect of discourse direction

12

Results

0%

20%

40%

60%

80%

100%

Normal Abnormal

% Interpretation

SourceGoal

Subjects: F(1,68)= 0.052p<0.820

Items: F(1,20)=0.111p<0.743

No effect of object type on pronoun interpretation

Page 13: Where to next?    Pronoun interpretation as a   side effect of discourse direction

13

Experiment 2: ‘What next?’ or ‘Why?’

(10) John handed a book to Bill. He ___________ .

Stimuli & Design: identical to Rohde et al. 2006

Predictions: “What next?” more Occasions Goal bias “Why?” more Explanations Source bias

Instructions: write continuations answering either “What happened next?” or “Why?”

Page 14: Where to next?    Pronoun interpretation as a   side effect of discourse direction

14

Methodology

Subjects: 42 monolingual English speakers Task: identical to Rohde et al. 2006 (w/instructions) Stimuli: identical to Rohde et al. 2006 Evaluation: judges assess coherence/interpretation

Analysis: Effect of between-subject factor of Instruction Type

on coherence distribution & pronoun interpretation

Page 15: Where to next?    Pronoun interpretation as a   side effect of discourse direction

15

0%

20%

40%

60%

80%

100%

What next? Why?

% Coherence

ExpElabV-EResOccPar

Results

Coherence varies w/instruction (p<0.0001)

Goal Source

0.81

0.811.00

0.110.28

0.46

Consistent prob(Source|coh)Exp 2Rohde et al.Coherence

Elaboration

0.87Violated-expectation

0.75Explanation0.99

0.16Result0.20Occasion

0.45Parallel

Page 16: Where to next?    Pronoun interpretation as a   side effect of discourse direction

16

Results

Effect of Instruction type on pronoun interpretation

0%

20%

40%

60%

80%

100%

What next? Why?

% Interpretation

SourceGoal

F(1,20)=52.672p<0.0001

Page 17: Where to next?    Pronoun interpretation as a   side effect of discourse direction

17

Predicting pronoun interpretation

=%Exp * p(SR |Exp) + %Elab* p(SR |Elab) + %V - E * p(SR | V - E) +

%Occ* p(SR |Occ) + %Res * p(SR |Res) + %Par* p(SR |Par)

p(Source)Coherence

Elaboration0.81

0.820.99

V-E

Explanation

0.050.17

ResultOccasion

0.45Parallel

Predict % Source Resolutions in Exp 2 using: Exp2 coherence breakdown Exp1 conditional probabilities

(1) %SR =

* p(SR |Coh)

coh∑

%coh

Page 18: Where to next?    Pronoun interpretation as a   side effect of discourse direction

18

Capturing subject variation

0%

20%

40%

60%

80%

100%

1 6 11 16 21 26 31 36 41 Participant

Observed%Source

Predicted%Sourceusing (1)

linear regression R2=0.604 F(1,40)=61.097, p<0.0001

* p(SR |Coh)

coh∑

(1) %SRi =

%cohi for subject i in Exp 2

“What next” “Why”

Page 19: Where to next?    Pronoun interpretation as a   side effect of discourse direction

19

R2=0.604, F(1,40)=61.097*Exp1: average across verbal aspects & object types

R2=0.586, F(1,40)=51.165*Exp1: imp, abnormal objects

R2=0.561, F(1,40)=51.165*Exp1: perf, abnormal objects

R2=0.627, F(1,40)=67.371*Exp1: imp, normal objects

R2=0.606, F(1,40)=61.612*Exp1: perf, normal objects

R2 value/ANOVAConditional Probability Estimator

* Indicates p<0.0001

Consistency of biases across conditions

Page 20: Where to next?    Pronoun interpretation as a   side effect of discourse direction

20

Shift coherence Shift pronoun interpretation No model relying only on surface-level cues can

account for observed variation, since stimuli were near-identical (Exp 1) or identical (Exp 2)

Need richer models incorporating discourse-level factors (see Wolf et al. 2004; Kertz et al. 2006)

Summary

Page 21: Where to next?    Pronoun interpretation as a   side effect of discourse direction

21

lowhigh Relative clause attachment ambiguity

What else can discourse do for you?

(11) Beth babysits the children of the musician who ____

Proposal: try to shift RC attachment using verbs that require Explanations and that attribute

cause to the referent occupying higher NP

musical prodigies themselves. arethe children

Function of a relative clause

(12) John despises the employee who is always late.

Implicit Causality (NP2 IC) verbs attribute cause to direct object

at the club downtown.playsthe musician

Page 22: Where to next?    Pronoun interpretation as a   side effect of discourse direction

22

Predictions & results

0%

20%

40%

60%

80%

100%

nonIC ICVerb Type

% Attachment

lowhigh

F(1,51)=31.082 p<0.0001

Further evidence that discourse influences interpretation

nonIC: Beth babysits the children of the musician who _____

0%

20%

40%

60%

80%

100%

nonIC ICVerb Type

% Coherence

ModExpElabOccResPar

p<0.0001

IC: Beth despises the children of the musician who ______

playsat the club downtown.

the musician low

and yell during rehearsals. screamthe children

high

Page 23: Where to next?    Pronoun interpretation as a   side effect of discourse direction

23

References

Arnold, J. E. (2001) The effects of thematic roles on pronoun use and frequency of reference. Discourse Processes, 31(2): 137-162.

Chambers, G. C. & Smyth, R. (1998) Structural parallelism and discourse coherence: A test of Centering Theory. Journal of Memory and Language, 39: 593-608.

Crawley, R., Stevenson, R., & Kleinman, D. (1990) The use of heuristic strategies in the interpretation of pronouns. Journal of Psycholinguistic Research, 4: 245–264.

Kameyama, M. (1996) Indefeasible semantics and defeasible pragmatics. In M. Kanazawa, C. Pinon, and H. de Swart, editors, Quantifiers, Deduction, and Context. CSLI Stanford, pp. 111-138.

Hobbs, J. R. (1979) Coherence and coreference, Cognitive Science, 3:67-90. Hobbs, J. R. (1990) Literature and Cognition. CSLI Lecture Notes 21. Stanford, CA. Kehler, A. (2002) Coherence, reference, and the theory of grammar. CSLI Publications, Stanford, CA.Kertz, L., Kehler, A., & Elman, J. (2006) Grammatical and Coherence-Based Factors in Pronoun

Interpretation. 28th Annual Conference of the Cognitive Science Society, Vancouver, July 2006.Moens, M. & Steedman, M. (1988) Temporal ontology and temporal reference. Computational

Linguistics 14(2):15-28.Smyth, R. H. (1994) Grammatical determinants of ambiguous pronoun resolution. Journal of

Psycholinguistic Research, 23: 197-229.Stevenson, R., Crawley R., & Kleinman D. (1994) Thematic roles, focusing and the representation of

events. Language and Cognitive Processes, 9:519–548.Wolf, F., Gibson, E. & Desmet, T. (2004) Coherence and pronoun resolution. Language and Cognitive

Processes, 19(6): 665-675

Page 24: Where to next?    Pronoun interpretation as a   side effect of discourse direction

24

Variation by instruction and aspect

(17) John was giving a book to Bill. He ___________ .(16) John gave a book to Bill. He ___________ .

…“What happened next?”

Interpretation (instr x aspect interaction p<0.0001)

0%

20%

40%

60%

80%

100%

whatnext?

why? whatnext?

why?

perfective imperfective

SourceGoal

0%

100%

whatnext?

why? whatnext?

why?

perfective imperfective

SourceGoal

Coherence (instr x aspect interaction p<0.0001)

0%

20%

40%

60%

80%

100%

Whatnext?

Why? Whatnext?

Why?

Perfective Imperfective

ParV-EElabExpResOcc0%

100%

Whatnext?

Why? Whatnext?

Why?

Perfective Imperfective

ParV-EElabExpResOcc

Page 25: Where to next?    Pronoun interpretation as a   side effect of discourse direction

25

Discourse coherence effectsPerfective Context Sentences

0

50

100

150

200

OccasionElaborationExplanationViolated-Exp

Result Parallel

Count

SourceGoal

Imperfective Context Sentences

0

50

100

150

200

OccasionElaborationExplanation

Result

Violated-ExpParallel

Count

SourceGoal