Learning When and What to Quote: A Quotation Recommender System with Mutual Promotion of Recommendation and Generation

Lingzhi Wang, Xingshan Zeng, Kam-Fai Wong

Abstract
This work extends the current quotation recommendation task to a more realistic quotation recommender system that learns to predict when to quote and what to quote jointly. The system consists of three modules (tasks), a prediction module to predict whether to quote given conversation contexts, a recommendation module to recommend suitable quotations and a generation module generating quotations or sentences in ordinary language to continue the conversation. We benchmark several competitive models for the two newly introduced tasks (i.e., when-to-quote and what-to-continue). For quotation recommendation, compared with previous work that is either generation-based or ranking-based recommendation, we propose a novel framework with mutual promotion of generation module and ranking-based recommendation module. Experiments show that our framework achieves significantly better performance than baselines on two datasets. Further experiments and analyses validate the effectiveness of the proposed mechanisms and get a better understanding of the quotation recommendation task.
Anthology ID:
2022.findings-emnlp.225
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2022
Month:
December
Year:
2022
Address:
Abu Dhabi, United Arab Emirates
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3094–3105
Language:
URL:
https://aclanthology.org/2022.findings-emnlp.225
DOI:
10.18653/v1/2022.findings-emnlp.225
Bibkey:
Cite (ACL):
Lingzhi Wang, Xingshan Zeng, and Kam-Fai Wong. 2022. Learning When and What to Quote: A Quotation Recommender System with Mutual Promotion of Recommendation and Generation. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 3094–3105, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
Cite (Informal):
Learning When and What to Quote: A Quotation Recommender System with Mutual Promotion of Recommendation and Generation (Wang et al., Findings 2022)
Copy Citation:
PDF:
https://preview.aclanthology.org/remove-xml-comments/2022.findings-emnlp.225.pdf
PDF Search