Publications

カンファレンス (国際) Cross-Domain Rating Prediction for Market Development

Yuki Otsuka (Kyoto univ.), Makoto P. Kato (Univ. Tsukuba), Sumio Fujita, Masatoshi Yoshikawa (Kyoto Univ.)

The 2020 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT'20)

2020.12.14

Most of the review-based prediction tasks are designed for an in-domain setting, in which a review score, sentiment, or sales in a domain is predicted by reviews in the same domain. In this paper, we address the rating prediction problem in a cross-domain setting, i.e. rating prediction of an item in a domain where no review has been given for the item yet, based on reviews provided in a different domain. For example, we predict the rating of a US product in the UK market, which has never been sold in the UK, in order to decide whether to start to sell the item in the UK. For effective cross-domain rating prediction, we propose a multitask learning model that predicts both the review rating and review contents simultaneously, and found that it was effective in two e-commerce datasets from Amazon and Yahoo. Furthermore, our experimental results revealed categories and cases in which the cross-domain rating prediction was successful.