Entrainable Neural Conversation Model Based on Reinforcement Learning
Entrainable Neural Conversation Model Based on Reinforcement Learning
复制标题
DOI：
10.1109/access.2020.3027099
复制
发表时间：
2020-09
期刊：
影响因子：
3.9
通讯作者：
Seiya Kawano;M. Mizukami;Koichiro Yoshino;Satoshi Nakamura
中科院分区：
文献类型：
--
作者：
Seiya Kawano;M. Mizukami;Koichiro Yoshino;Satoshi Nakamura
对话中词语的同步，称为夹带，通常在人与人的对话中观察到。夹带与对话的成功、自然度和参与度具有高度相关性。在本文中，我们根据语义空间中的单词相似度定义夹带分数来评估系统生成的夹带。我们使用强化学习针对夹带分数优化了神经对话模型，以便系统可以控制系统响应的夹带程度。实验结果表明，所提出的可夹带神经对话模型产生了与传统模型相当或更自然的响应，并且令人满意地控制了所生成响应的夹带程度。
The synchronization of words in conversation, called entrainment, is generally observed in human-human conversations. Entrainment has a high correlation with dialogue success, naturalness, and engagement. In this article, we define entrainment scores based on the word similarities in semantic space to evaluate the entrainment of system generation. We optimized a neural conversation model to the entrainment scores using reinforcement learning so that the system can control the degree of entrainment of the system response. Experimental results showed that the proposed entrainable neural conversation model generated comparable or more natural responses than conventional models and satisfactorily controlled the degree of entrainment of the generated responses.