Cross-Lingual Transfer of Cognitive Processing Complexity

Charlotte Pouw; Nora Hollenstein; Lisa Beinborn

doi:10.18653/v1/2023.findings-eacl.49

Cross-Lingual Transfer of Cognitive Processing Complexity

Charlotte Pouw, Nora Hollenstein, Lisa Beinborn

Research output: Chapter in Book / Report / Conference proceeding › Conference contribution › Academic › peer-review

Abstract

When humans read a text, their eye movements are influenced by the structural complexity of the input sentences. This cognitive phenomenon holds across languages and recent studies indicate that multilingual language models utilize structural similarities between languages to facilitate cross-lingual transfer. We use sentence-level eye-tracking patterns as a cognitive indicator for structural complexity and show that the multilingual model XLM-RoBERTa can successfully predict varied patterns for 13 typologically diverse languages, despite being fine-tuned only on English data. We quantify the sensitivity of the model to structural complexity and distinguish a range of complexity characteristics. Our results indicate that the model develops a meaningful bias towards sentence length but also integrates cross-lingual differences. We conduct a control experiment with randomized word order and find that the model seems to additionally capture more complex structural information.

Original language	English
Title of host publication	Findings of the Association for Computational Linguistics: EACL 2023
Subtitle of host publication	[Proceedings]
Publisher	Association for Computational Linguistics (ACL)
Pages	643-657
Number of pages	15
ISBN (Electronic)	9781959429470
DOIs	https://doi.org/10.18653/v1/2023.findings-eacl.49
Publication status	Published - 2023
Event	17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023 - Findings of EACL 2023 - Dubrovnik, Croatia Duration: 2 May 2023 → 6 May 2023

Conference

Conference	17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023 - Findings of EACL 2023
Country/Territory	Croatia
City	Dubrovnik
Period	2/05/23 → 6/05/23

Bibliographical note

Funding Information:
We thank the anonymous reviewers for their insightful feedback. L. Beinborn’s research was supported by the Dutch National Science Organisation (NWO) through the projects CLARIAHPLUS (CP-W6-19-005) and VENI (Vl.Veni.211C.039).

Publisher Copyright:
© 2023 Association for Computational Linguistics.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.18653/v1/2023.findings-eacl.49Licence: CC BY

https://aclanthology.org/2023.findings-eacl.49/

Persistent URL (handle)

Persistent link

Cite this

@inproceedings{787f7a8c486c470cb0845c533af83d50,

title = "Cross-Lingual Transfer of Cognitive Processing Complexity",

abstract = "When humans read a text, their eye movements are influenced by the structural complexity of the input sentences. This cognitive phenomenon holds across languages and recent studies indicate that multilingual language models utilize structural similarities between languages to facilitate cross-lingual transfer. We use sentence-level eye-tracking patterns as a cognitive indicator for structural complexity and show that the multilingual model XLM-RoBERTa can successfully predict varied patterns for 13 typologically diverse languages, despite being fine-tuned only on English data. We quantify the sensitivity of the model to structural complexity and distinguish a range of complexity characteristics. Our results indicate that the model develops a meaningful bias towards sentence length but also integrates cross-lingual differences. We conduct a control experiment with randomized word order and find that the model seems to additionally capture more complex structural information.",

author = "Charlotte Pouw and Nora Hollenstein and Lisa Beinborn",

note = "Funding Information: We thank the anonymous reviewers for their insightful feedback. L. Beinborn{\textquoteright}s research was supported by the Dutch National Science Organisation (NWO) through the projects CLARIAHPLUS (CP-W6-19-005) and VENI (Vl.Veni.211C.039). Publisher Copyright: {\textcopyright} 2023 Association for Computational Linguistics.; 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023 - Findings of EACL 2023 ; Conference date: 02-05-2023 Through 06-05-2023",

year = "2023",

doi = "10.18653/v1/2023.findings-eacl.49",

language = "English",

pages = "643--657",

booktitle = "Findings of the Association for Computational Linguistics: EACL 2023",

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

address = "United States",

}

Pouw, C, Hollenstein, N & Beinborn, L 2023, Cross-Lingual Transfer of Cognitive Processing Complexity. in Findings of the Association for Computational Linguistics: EACL 2023: [Proceedings]. Association for Computational Linguistics (ACL), pp. 643-657, 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023 - Findings of EACL 2023, Dubrovnik, Croatia, 2/05/23. https://doi.org/10.18653/v1/2023.findings-eacl.49

Cross-Lingual Transfer of Cognitive Processing Complexity. / Pouw, Charlotte; Hollenstein, Nora; Beinborn, Lisa.
Findings of the Association for Computational Linguistics: EACL 2023: [Proceedings]. Association for Computational Linguistics (ACL), 2023. p. 643-657.

Research output: Chapter in Book / Report / Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Cross-Lingual Transfer of Cognitive Processing Complexity

AU - Pouw, Charlotte

AU - Hollenstein, Nora

AU - Beinborn, Lisa

N1 - Funding Information: We thank the anonymous reviewers for their insightful feedback. L. Beinborn’s research was supported by the Dutch National Science Organisation (NWO) through the projects CLARIAHPLUS (CP-W6-19-005) and VENI (Vl.Veni.211C.039). Publisher Copyright: © 2023 Association for Computational Linguistics.

PY - 2023

Y1 - 2023

N2 - When humans read a text, their eye movements are influenced by the structural complexity of the input sentences. This cognitive phenomenon holds across languages and recent studies indicate that multilingual language models utilize structural similarities between languages to facilitate cross-lingual transfer. We use sentence-level eye-tracking patterns as a cognitive indicator for structural complexity and show that the multilingual model XLM-RoBERTa can successfully predict varied patterns for 13 typologically diverse languages, despite being fine-tuned only on English data. We quantify the sensitivity of the model to structural complexity and distinguish a range of complexity characteristics. Our results indicate that the model develops a meaningful bias towards sentence length but also integrates cross-lingual differences. We conduct a control experiment with randomized word order and find that the model seems to additionally capture more complex structural information.

AB - When humans read a text, their eye movements are influenced by the structural complexity of the input sentences. This cognitive phenomenon holds across languages and recent studies indicate that multilingual language models utilize structural similarities between languages to facilitate cross-lingual transfer. We use sentence-level eye-tracking patterns as a cognitive indicator for structural complexity and show that the multilingual model XLM-RoBERTa can successfully predict varied patterns for 13 typologically diverse languages, despite being fine-tuned only on English data. We quantify the sensitivity of the model to structural complexity and distinguish a range of complexity characteristics. Our results indicate that the model develops a meaningful bias towards sentence length but also integrates cross-lingual differences. We conduct a control experiment with randomized word order and find that the model seems to additionally capture more complex structural information.

UR - http://www.scopus.com/inward/record.url?scp=85159858247&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85159858247&partnerID=8YFLogxK

U2 - 10.18653/v1/2023.findings-eacl.49

DO - 10.18653/v1/2023.findings-eacl.49

M3 - Conference contribution

AN - SCOPUS:85159858247

SP - 643

EP - 657

BT - Findings of the Association for Computational Linguistics: EACL 2023

PB - Association for Computational Linguistics (ACL)

T2 - 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023 - Findings of EACL 2023

Y2 - 2 May 2023 through 6 May 2023

ER -

Cross-Lingual Transfer of Cognitive Processing Complexity

Abstract

Conference

Bibliographical note

UN SDGs

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this