| 05 |
Dialog processing techniques for reading the situationMultimodal situation recognition for everyday conversations 
|
|---|
Dialogue systems must correctly understand the dialogue situations to participate in everyday conversation naturally. We introduce two studies for understanding situations of everyday conversations. In the first research, we conducted a factor analysis to investigate the characteristics of everyday conversations. From the analysis, we clarified that our daily conversation could be explained by the combination of seven factors related to the dialogue purpose and manner. In the second research, we proposed a method to recognize dialogue situations, including the place of the conversation and the relationships among the participants by using multimodal information. In recent years, the naturalness of responses in dialogue systems has dramatically improved. Integrating multimodal information processing allows the system to have social conversations with humans in everyday situations. By developing our techniques, we aim to realize a future in which it will be commonplace to converse with dialogue systems.
[1] Y. Chiba and R. Higashinaka, “Dialogue situation recognition for everyday conversation using multimodal information,” in Proc. INTERSPEECH, pp. 241-245, 2021.
[2] Y. Chiba and R. Higashinaka, “Analyzing variations of everyday Japanese conversations based on semantic labels of functional expressions,” ACM Transactions on Asian and Low-Resource Language Information Processing, Vol. 22, No. 2, pp. 1-26, 2022.
Yuya Chiba
Interaction Research Group, Innovative Communication Laboratory