Overview of the NLPCC 2023 Shared Task 9: User Feedback Prediction and Response Generation

Hanlin Teng, Hongda Sun, Wei Liu, Shuang Dong, Rui Yan, Jian Luan, Bin Wang

Abstract: This paper presents an overview of user feedback prediction and response generation in the NLPCC 2023 shared task. We focus on how to utilize feedback data of user likes and dislikes to guide conversation response generation. The goal of this task is to predict accurate user preference and improve response quality to increase user likes. Participants need to integrate preference information into their models to generate responses that align with the user needs. In this paper, we summarize the key components of this task, including task description, dataset, evaluation metrics, participant methods, and final results. We also highlight the potential applications of incorporating like and dislike data in conversation generation.