Host: The Japanese Society for Artificial Intelligence
Name : 34th Annual Conference, 2020
Number : 34
Location : Online
Date : June 09, 2020 - June 12, 2020
People in the business companies and academic fields work in cooperation with others rather than working alone. They may discuss their progress with others like co-workers and supervisors to obtain more achievements. Sometimes people may feel that such discussions are not conducted well. However, people do not evaluate the quality of each discussion every time because it is tough work for them; they usually do not have enough time for that. As evaluating discussions, people might look back on their discussions and conduct improved discussions from the next time. This paper proposes an evaluation method for one-to-one research discussions. The method makes a model of discussion process with transitions of utterance types. The labels of utterance types are put on each utterance in a discussion text by manually. As calculating the transition probabilities between two labels, a matrix of transition probabilities is obtained. The transitions of labels with high probabilities are extracted from good discussions and connected to obtain a model of the discussion process. The model can be used for evaluation of new discussion. We applied the proposed method to the discussion texts and found that the obtained process model from good discussions was different from that from not-good discussions.