A Method for Using DUs in Spontaneous Conversation

Using Discounted Utterances (DUs) in spontaneous conversation by applying text mining technology, extraction, and evaluation, we focused on DUs where values were buried in previous conversations. We discovered DU potentials by reconsidering them through human-computer interaction. Onlinechat experiments clarified DU features and demonstrated our system’s importance. We found DUs involving (1) experiences shared by the subjects, (2) subjects’ unique experiences, concerns, or beliefs, and (3) apparent unimportance or unrecognized potential. Results of the experiments showed our evaluation method to be appropriate for calculating DU importance when DUs involving (3) were valued significantly lower than (1) and (2). Experiments also suggested that most DUs extracted by the system were not indeed completely ignored but included subjects’ unique stories involving main contexts. Such stories were based on subjects’ unique experiences and may be useful for helping subjects’ metacognition. The system may also enable nonsubjects to infer subjects and their thinking.