Overall Rating Prediction from Review Texts using Category-oriented Japanese Sentiment Polarity Dictionary

Abstract

Hotel booking sites provide evaluations, including textual reviews and numerical ratings by hotel guests. However, some evaluations are without numerical ratings, and there are some inconsistent evaluations between textual reviews and numerical ratings (e.g., a positive review text is posted along with low ratings). Such evaluations may confuse site users. To resolve such problems, we propose a high-accuracy method to predict an overall numerical rating from a textual review. Our new idea is to use Category-oriented Sentiment Polarity Dictionaries (CSPD), which are automatically compiled for each category using a Rakuten Travel review database. The CSPD gives the sentiment polarity value (i.e., the positivity/negativity value) for each sentiment word for each category. Our proposed method first predicts category ratings from its BERT vector and the CSPD. After that, based on the predicted category ratings and the BERT vector, our method predicts the overall rating. Our experimental results show that our method attains higher accuracy than using only BERT vectors.