Dialog Act Classification for Vietnamese Spoken Text

Thi-Lan Ngo, Thi Bich Ngoc Doan, Thi Lan Phuong Ngo

Systems, which use the conversational interface to interact with users such as chat-bots, virtual personal assistants, recommendation systems and automatic customer care systems and so on, are getting popular in our life. A significant challenge in designing and building those systems is how to effectively determine the user intents from the user’s speech interactions. In particular, determining dialog act is the first step in determining user intent. Dialog act recognition has widely studied in many different languages but in Vietnamese, there are few studies. In this paper, we present an attempt on dialog act recognition for Vietnamese conversational text. We adopt a machine learning approach by using maximum entropy model on Vietnamese conversational dataset labelled dialog act based on ISO 24617-2 standard. The achieved result is 70.80% that satisfy for practical applications.