Verification of Evaluation Model for Speech Proactiveness in Foreign Language Conversations

The purpose of this study is to estimate the speech proactivity of Japanese foreign-language learners in foreign-language communication. First, we focused on intonation change, utterance time ratio, and average voice volume as parameters for evaluating speech proactivity. We assumed that higher values indicate the impression of proactivity. Based on this prediction, we defined a linear model to predict speech proactivity using these evaluation parameters as independent variables. We implemented the speech proactivity evaluation system that applied this mathematical model using Pure Data and Unity. To verify the validity and suitability of the evaluation parameters for the prepared speech proactivity evaluation model, was conducted an experiment with 24 Japanese students. The participants listened to short English conversations between foreigners and Japanese people and gave their impressions of Japanese-speaking attitudes. Multiple regression analysis was conducted using the degree of intonation change, utterance time ratio, and average voice volume obtained by the system as independent variables, and the impression of proactivity was obtained through the experiment as a dependent variable. The analysis yielded a significant regression model. We found that the degree of intonation change and the ratio of speech time had a positive causal relationship with the impression of proactiveness. The above shows that the parameter selection of each evaluation parameter of the speech proactivity evaluation model is generally appropriate.