Recognition of Personality Traits using Word Vector from Reflective Context

摘要

Predicting personality is meaningful for many social applications that target humans. In this work we proposed a way to model the user's personality with a small number of contexts based on personal SNS post data. We compared and analyzed various combination of word vector and classifier to optimize performance. We find that our model achieves f1-scores 0.72 and 0.74 in unimodal and multimodal case respectively for Big-5 personality traits. We are planning to develop a real time personality recognizer that operates with utterance in the human-robot interaction situation.