Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue

Aron Molnar; Jaap Jumelet; Mario Giulianelli; Arabella Sinclair

doi:10.18653/v1/2023.conll-1.18

Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue

Aron Molnar, Jaap Jumelet, Mario Giulianelli, Arabella Sinclair

University of Amsterdam

Research output: Chapter in Book/Report/Conference proceeding › Published conference contribution

Abstract

Language models are often used as the backbone of modern dialogue systems. These models are pre-trained on large amounts of written fluent language. Repetition is typically penalised when evaluating language model generations. However, it is a key component of dialogue. Humans use local and partner specific repetitions; these are preferred by human users and lead to more successful communication in dialogue. In this study, we evaluate (a) whether language models produce human-like levels of repetition in dialogue, and (b) what are the processing mechanisms related to lexical re-use they use during comprehension. We believe that such joint analysis of model production and comprehension behaviour can inform the development of cognitively inspired dialogue generation systems.

Original language	English
Title of host publication	CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings
Editors	Jing Jiang, David Reitter, Shumin Deng
Publisher	Association for Computational Linguistics (ACL)
Pages	254-273
Number of pages	20
ISBN (Electronic)	9798891760394
DOIs	https://doi.org/10.18653/v1/2023.conll-1.18
Publication status	Published - Dec 2023
Event	27th Conference on Computational Natural Language Learning, CoNLL 2023 - Singapore, Singapore Duration: 6 Dec 2023 → 7 Dec 2023

Conference

Conference	27th Conference on Computational Natural Language Learning, CoNLL 2023
Country/Territory	Singapore
City	Singapore
Period	6/12/23 → 7/12/23

Bibliographical note

Funding Information:
We would like to thank the anonymous reviewers for their thoughtful and useful reviews and comments. We also wish to thank Ehud Reiter for his useful comments on this work at an early stage. MG is supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 819455).

Access to Document

10.18653/v1/2023.conll-1.18Licence: CC BY

Molnar_etal_CONLL_Attribution_and_Alignment_VOR
https://creativecommons.org/licenses/by/4.0/
Final published version, 4.59 MBLicence: CC BY

Cite this

Molnar, A., Jumelet, J., Giulianelli, M., & Sinclair, A. (2023). Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue. In J. Jiang, D. Reitter, & S. Deng (Eds.), CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings (pp. 254-273). Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/2023.conll-1.18

Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue. / Molnar, Aron; Jumelet, Jaap; Giulianelli, Mario et al.
CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings. ed. / Jing Jiang; David Reitter; Shumin Deng. Association for Computational Linguistics (ACL), 2023. p. 254-273.

Research output: Chapter in Book/Report/Conference proceeding › Published conference contribution

Molnar, A, Jumelet, J, Giulianelli, M & Sinclair, A 2023, Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue. in J Jiang, D Reitter & S Deng (eds), CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings. Association for Computational Linguistics (ACL), pp. 254-273, 27th Conference on Computational Natural Language Learning, CoNLL 2023, Singapore, Singapore, 6/12/23. https://doi.org/10.18653/v1/2023.conll-1.18

Molnar A, Jumelet J, Giulianelli M, Sinclair A. Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue. In Jiang J, Reitter D, Deng S, editors, CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings. Association for Computational Linguistics (ACL). 2023. p. 254-273 doi: 10.18653/v1/2023.conll-1.18

Molnar, Aron ; Jumelet, Jaap ; Giulianelli, Mario et al. / Attribution and Alignment : Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue. CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings. editor / Jing Jiang ; David Reitter ; Shumin Deng. Association for Computational Linguistics (ACL), 2023. pp. 254-273

@inproceedings{b032bf5700fb470dbc17d7def6a3434b,

title = "Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue",

abstract = "Language models are often used as the backbone of modern dialogue systems. These models are pre-trained on large amounts of written fluent language. Repetition is typically penalised when evaluating language model generations. However, it is a key component of dialogue. Humans use local and partner specific repetitions; these are preferred by human users and lead to more successful communication in dialogue. In this study, we evaluate (a) whether language models produce human-like levels of repetition in dialogue, and (b) what are the processing mechanisms related to lexical re-use they use during comprehension. We believe that such joint analysis of model production and comprehension behaviour can inform the development of cognitively inspired dialogue generation systems.",

author = "Aron Molnar and Jaap Jumelet and Mario Giulianelli and Arabella Sinclair",

note = "Funding Information: We would like to thank the anonymous reviewers for their thoughtful and useful reviews and comments. We also wish to thank Ehud Reiter for his useful comments on this work at an early stage. MG is supported by the European Research Council (ERC) under the European Union{\textquoteright}s Horizon 2020 research and innovation programme (grant agreement No. 819455). ; 27th Conference on Computational Natural Language Learning, CoNLL 2023 ; Conference date: 06-12-2023 Through 07-12-2023",

year = "2023",

month = dec,

doi = "10.18653/v1/2023.conll-1.18",

language = "English",

pages = "254--273",

editor = "Jing Jiang and David Reitter and Shumin Deng",

booktitle = "CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings",

publisher = "Association for Computational Linguistics (ACL)",

}

TY - GEN

T1 - Attribution and Alignment

T2 - 27th Conference on Computational Natural Language Learning, CoNLL 2023

AU - Molnar, Aron

AU - Jumelet, Jaap

AU - Giulianelli, Mario

AU - Sinclair, Arabella

N1 - Funding Information: We would like to thank the anonymous reviewers for their thoughtful and useful reviews and comments. We also wish to thank Ehud Reiter for his useful comments on this work at an early stage. MG is supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 819455).

PY - 2023/12

Y1 - 2023/12

N2 - Language models are often used as the backbone of modern dialogue systems. These models are pre-trained on large amounts of written fluent language. Repetition is typically penalised when evaluating language model generations. However, it is a key component of dialogue. Humans use local and partner specific repetitions; these are preferred by human users and lead to more successful communication in dialogue. In this study, we evaluate (a) whether language models produce human-like levels of repetition in dialogue, and (b) what are the processing mechanisms related to lexical re-use they use during comprehension. We believe that such joint analysis of model production and comprehension behaviour can inform the development of cognitively inspired dialogue generation systems.

AB - Language models are often used as the backbone of modern dialogue systems. These models are pre-trained on large amounts of written fluent language. Repetition is typically penalised when evaluating language model generations. However, it is a key component of dialogue. Humans use local and partner specific repetitions; these are preferred by human users and lead to more successful communication in dialogue. In this study, we evaluate (a) whether language models produce human-like levels of repetition in dialogue, and (b) what are the processing mechanisms related to lexical re-use they use during comprehension. We believe that such joint analysis of model production and comprehension behaviour can inform the development of cognitively inspired dialogue generation systems.

UR - http://www.scopus.com/inward/record.url?scp=85184518153&partnerID=8YFLogxK

U2 - 10.18653/v1/2023.conll-1.18

DO - 10.18653/v1/2023.conll-1.18

M3 - Published conference contribution

AN - SCOPUS:85184518153

SP - 254

EP - 273

BT - CoNLL 2023 - 27th Conference on Computational Natural Language Learning, Proceedings

A2 - Jiang, Jing

A2 - Reitter, David

A2 - Deng, Shumin

PB - Association for Computational Linguistics (ACL)

Y2 - 6 December 2023 through 7 December 2023

ER -

Attribution and Alignment: Effects of Local Context Repetition on Utterance Production and Comprehension in Dialogue

Abstract

Conference

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this