Analysis of relations between spoken utterances and text messages in chat-augmented meetings

A chat-augmented meeting (a face-to-face meeting scenario where a text chat system is concomitantly used) has recently attracted a lot of attention, and many related research attempts have been made. We have been attempting to create a retrieval system for spoken utterances by using text messages in the text chat system as keywords for such a hybrid discussion. Toward this goal, we investigate the temporal relations between the spoken utterances and the text messages. We obtain a decision tree based on the J4.8 classifier and evaluate its performance. Although the performance of this decision tree is currently not so high, it still achieved a level that is promising for practical use after future improvements.