Scholarlink

<p>Conversational Recommender Systems (CRS) aim to understand user preferences through dialogue and provide appropriate recommendations. High-quality conversational context is crucial to improving CRS performance, but existing datasets often suffer from unnatural dialogue flow, simplistic interactions, limited scope, and unrealistic user preferences. Different from previous datasets, CeReal integrates user personality and rich metadata for realistic and personalized dialog. This study introduces CeReal, a high-quality CRS dataset in the movie domain, designed to address these challenges. CeReal leverages actual viewing history and ratings of users to assign personality types, allowing more natural and personalized dialogue flows. The dataset also provides LLMs with three key information types—User, Movie, and Personality information—facilitating rich interactions that reflect diverse scenarios and contextual details. Extensive experiments validate the high-quality conversational context in CeReal and demonstrate its effectiveness in enabling personality-based customized interactions, making it a valuable resource for advancing CRS development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CEREAL: personality-driven LLM-based conversational recommendation dataset with contextually-enriched and realistic user interactions

  • 期刊论文

  • 发表日期：

    2026年1月29日

查看全文
  • Jiyoon Lee,
  • Joonghoon Kim,
  • Pilsung Kang

摘要

Conversational Recommender Systems (CRS) aim to understand user preferences through dialogue and provide appropriate recommendations. High-quality conversational context is crucial to improving CRS performance, but existing datasets often suffer from unnatural dialogue flow, simplistic interactions, limited scope, and unrealistic user preferences. Different from previous datasets, CeReal integrates user personality and rich metadata for realistic and personalized dialog. This study introduces CeReal, a high-quality CRS dataset in the movie domain, designed to address these challenges. CeReal leverages actual viewing history and ratings of users to assign personality types, allowing more natural and personalized dialogue flows. The dataset also provides LLMs with three key information types—User, Movie, and Personality information—facilitating rich interactions that reflect diverse scenarios and contextual details. Extensive experiments validate the high-quality conversational context in CeReal and demonstrate its effectiveness in enabling personality-based customized interactions, making it a valuable resource for advancing CRS development.