Dialogue Acts Classification via RNNs with One- Versus-All Layers Considering Rare Utterances
Haruno Izumi, Shōhei Kato · 2019
For understanding the contents of a user's utterances, dialogue acts classification is introduced to chat systems. Because the number of utterances differs greatly in each dialogue act, classifiers tend to confuse dialogue acts that rarely appear in dialogue with frequent ones. This paper proposes Enhanced One-versus-All RNN (ENOVA RNN), which is a method to capture some features of dialogue acts that rarely appear in dialogues and discusses the effectiveness. In this study, it was confirmed that ENOVA RNN can improve rare dialogue acts classification performance keeping the overall quality of the performance.