Chinese Emotional Dialogue Response Generation via Reinforcement Learning
Rushi Lan,Jing Wang,Wenming Huang,Zhenrong Deng,Xiyan Sun,Zhuo Chen,Xiaonan Luo
DOI: https://doi.org/10.1145/3446390
IF: 5.3
2021-07-22
ACM Transactions on Internet Technology
Abstract:In an open-domain dialogue system, recognition and expression of emotions are the key factors for success. Most of the existing research related to Chinese dialogue systems aims at improving the quality of content but ignores the expression of human emotions. In this article, we propose a Chinese emotional dialogue response generation algorithm based on reinforcement learning that can generate responses not only according to content but also according to emotion. In the proposed method, a multi-emotion classification model is first used to add emotion labels to the corpus of post-response pairs. Then, with the help of reinforcement learning, the reward function is constructed based on two aspects, namely, emotion and content. Among the generated candidates, the system selects the one with long-term success as the best reply. At the same time, to avoid safe responses and diversify dialogue, a diversity beam search algorithm is applied in the decoding process. The comparative experiments demonstrate that the proposed model achieves satisfactory results according to both automatic and human evaluations.
computer science, information systems, software engineering