Multimodal Analysis of Client Persuasion in Consulting Interactions: Toward Understanding Successful Consulting

摘要

To analyze successful consulting processes using multimodal analysis, the aim of this research is to develop a model for recognizing when a client is persuaded by a consultant using multimodal features. These models enable us to analyze the utterances of highly skilled professional consultants in persuading clients. For this purpose, first, we collect a multimodal counseling interaction corpus including audio and spoken dialogue content (manual transcription) on dialogue sessions between a professional beauty counselor and five clients. Second, we developed a recognition model of persuasion labels using acoustic and linguistic features that are extracted from a multimodal corpus by training a machine learning model as a binary classification task. The experimental results show that the persuasion was 0.697 for accuracy and 0.661 for F1-score with bidirectional LSTM.