Ewha Womans University's Department of Data Science and commerce artificial intelligence startup AI.M announced on the 17th that their jointly developed conversational recommendation system research won the Best Paper Award at the '2025 Korea Computer Conference (KCC 2025)'.
This award-winning paper is titled *'A Study on Improving Interactive Recommendations by Reflecting User Preferences for Intrinsic and Extrinsic Product Characteristics'*, co-authored by Professor Choi Ye-rim (Department of Data Science, Ewha Womans University, CEO of AI.m) and researchers Choi Bo-hyun, Han Gyu-rim, and Park Ji-yoon from the Information Management Lab at Ewha Womans University.
The study presented a new approach to complement the existing limitations of conversational recommendation systems. The research team estimated product attribute preferences that were not explicitly stated in the conversation process based on the user's purchase history, and developed a personalized recommendation system that reflects both intrinsic attributes (e.g., color, size) and extrinsic attributes (e.g., popularity, review satisfaction). While existing systems perform recommendations based on directly mentioned information, this study is different in that it enables more precise preference analysis by utilizing user behavior data.
The experimental results showed that as the user's purchase history accumulated to a certain level, the recommendation performance improved, but after a certain number of purchases was exceeded, the performance tended to slightly decrease. The research team interpreted this as a phenomenon related to the user's 'diversity-seeking behavior.'
Professor Choi Ye-rim said, “This study presents a new methodology for reflecting extrinsic attributes that are difficult to directly confirm through conversation into recommendations,” and added, “This technology can be applied to AI.com’s commerce solution to provide a more sophisticated personalized shopping experience.”
- See more related articles
You must be logged in to post a comment.