Interest Estimation for a Personalized Museum Guide

Abstract

This paper propose a model that estimates the user's interest in artwork and the description sentences based on the user's profile and the history of the user's interest in previous interacted artworks in order to develop a museum guide system that provides highly personalized exhibit explanation for each user. A text corpus with annotations of the discourse structure, users' profiles, and interests in artwork and the description sentences was initially constructed to evaluate the proposed method. Experts annotated the discourse structure for the description texts of national treasures and important cultural properties. Users' profiles and interests in artwork and the description sentences were collected via crowdsourcing. Experiments using this corpus demonstrated that the interest estimation performance of the proposed model improves as the number of context artworks increases.