Detecting microstructures of conversations by using physical references: case study of poster presentations

The purpose of the work presented in this paper is reusing conversational archives for supporting knowledge circulation in the real world. So far, we have proposed a method to give structures to conversations by analyzing gaze and utterance data from an interaction corpus, which is a semi-structured set of a large amount of interaction data collected by various sensors. In this paper, we give more detailed structures to conversations by analyzing poster touch data. We devise a poster touch capturing method and investigate the correlation between poster touch and conversation structure, that is, a topic of conversation has a relation to a sub-theme in a poster that an exhibitor touched at the time when it was made. By analyzing the relation between conversations and poster touch data, we can detect the transition of topics and divide a long conversation that is composed of many topics into some microstructured conversations that are composed of a single topic or context.