Dialogue-based tourist spot recommendation system using word embeddings

Abstract

In this study, we propose a system that recommends tourist spots matching the user’s preferences through dialogue. We collect texts about tourist spots from Japanese Wikipedia and tourism information websites to create a corpus, with which we create embeddings for tourist spots using word2vec. When a user inputs desired tourism elements through dialogue, the system recommends tourist spots that are most similar to these elements based on embeddings. To confirm the effectiveness of the proposed method, we report the results of an experiment comparing the effectiveness of it with a word frequency-based recommendation method.