국희진 석사과정, 김준영 석사과정, 박성민 석박통합과정 NAACL 2025 국제 학술대회 논문 채택

23 Jan 2025

DIAL 연구실 소속 인공지능학과 국희진(석사과정, 공동 제1저자) 학생, 김준영(석사과정, 공동 제1저자) 학생, 박성민(석·박통합과정, 제3저자) 학생, 이종욱(교신저자) 교수가 참여한 “Empowering Retrieval-based Conversational Recommendation with Ambivalent User Preferences” 논문이 자연어처리 분야 최우수 국제 학술대회인 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL)에 최종 게재가 승인되었으며 오는 4월에 발표될 예정입니다.

Abstract

Conversational recommender systems (CRSs) are designed to suggest the target item that the user is likely to prefer through multi-turn conversations. Recent studies stress that capturing sentiments in user conversations improves recommendation accuracy. However, they employ a single user representation, which fails to distinguish between conflicting user intentions, such as likes and dislikes, potentially leading to suboptimal performance. To this end, we propose a novel conversational recommender method, AMbivalent user Preference expansion and LEarning (AMPLE). Firstly, AMPLE enhances the user's potential preferences through ambivalent preference expansion by leveraging the reasoning capacity of the LLMs. Based on the potential preference, AMPLE explicitly differentiates the ambivalent preferences and leverages them into the recommendation process via preference-aware learning. Extensive experiments show that AMPLE significantly outperforms existing methods across benchmarks, improving up to 99.72% in Recall@10.