A Real-Time Combination of Text-Mining Methods for Flexible Movie Recommendation in Human Robot Interaction

Abstract

For more realistic human-robot interaction, a robot should be able to flexibly respond to human's linguistic expressions that are not predefined in situations of face-to-face communication. However, most robots currently employ a limited response method in which they only react when the human speaks predefined words or sentences in a dictionary. This has been regarded as a limitation to the practical application of robots in real life. In this study, a text mining-based recommendation method was developed for robots to understand the meaning of exceptional human speech and obtain knowledge by using many external corpora with related data or knowledge based on the content of human speech. Tf-idf and LDA are combined to increase the recommendation accuracy.