Domain action classification using a maximum entropy model in a schedule management domain

Abstract

To generate correct reactions, a dialogue system should identify domain actions indicated by users' utterances because speaker intentions can be captured by the domain actions. In this paper, we propose a domain action classification model to determine speech acts (general intentions) and concept sequences (semantic focuses) at the same time in a schedule management domain. To avoid biased learning problems, the proposed model uses low-level linguistic features and filters out uninformative features using statistic. Then, the proposed model determines domain actions using a maximum entropy model. In the experiment, the proposed model showed better performances than previous works in speech act classification. In addition, the proposed model showed high performances in concept sequence classification. Based on these experimental results, we believe that the proposed model will be more helpful to a dialogue system than previous speech act classification models because it can return speech acts and concept sequences at the same time on the same framework.