Assessing Word Importance Using Models Trained for Semantic Tasks

Dávid Javorský, Ondřej Bojar, François Yvon


Abstract
Many NLP tasks require to automatically identify the most significant words in a text. In this work, we derive word significance from models trained to solve semantic task: Natural Language Inference and Paraphrase Identification. Using an attribution method aimed to explain the predictions of these models, we derive importance scores for each input token. We evaluate their relevance using a so-called cross-task evaluation: Analyzing the performance of one model on an input masked according to the other model’s weight, we show that our method is robust with respect to the choice of the initial task. Additionally, we investigate the scores from the syntax point of view and observe interesting patterns, e.g. words closer to the root of a syntactic tree receive higher importance scores. Altogether, these observations suggest that our method can be used to identify important words in sentences without any explicit word importance labeling in training.
Anthology ID:
2023.findings-acl.563
Volume:
Findings of the Association for Computational Linguistics: ACL 2023
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
8846–8856
Language:
URL:
https://aclanthology.org/2023.findings-acl.563
DOI:
10.18653/v1/2023.findings-acl.563
Bibkey:
Cite (ACL):
Dávid Javorský, Ondřej Bojar, and François Yvon. 2023. Assessing Word Importance Using Models Trained for Semantic Tasks. In Findings of the Association for Computational Linguistics: ACL 2023, pages 8846–8856, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
Assessing Word Importance Using Models Trained for Semantic Tasks (Javorský et al., Findings 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.findings-acl.563.pdf