When Less Is More: Using Less Context Information to Generate Better Utterances in Group Conversations

Abstract

Previous research on dialogue systems generally focuses on the conversation between two participants. Yet, group conversations which involve more than two participants within one session bring up a more complicated situation. The scenario is real such as meetings or online chatting rooms. Learning to converse in groups is challenging due to different interaction patterns among users when they exchange messages with each other. Group conversations are structure-aware while the structure results from different interactions among different users. In this paper, we have an interesting observation that fewer contexts can lead to better performance by tackling the structure of group conversations. We conduct experiments on the public Ubuntu Multi-Party Conversation Corpus and the experiment results demonstrate that our model outperforms baselines.