情報処理学会 第76回全国大会講演要旨

In cuisine recommender service, on-line user review is an important data source avoiding a cold-start problem. Cuisine-domain named entity recognition can be used as an entrance to comprehend the semantic information of reviews. This paper describes a supervised approach recognizing Japanese dish name entity (DNE) from on-line reviews of Japanese cuisine website. In the first stage, this work adopts tweets as the data source to construct seed dictionary of dish name elements through semantic rules and uses Bayesian posterior to remove noise. Next stage, we map first-stage dictionary as a non-local feature into Condition Random Field to extract the dish name. This method can automatically add new dish name elements into the seed dictionary by iteration during the recognition proceeding. By using 10-fold testing, experiment results show our method can reach 84.38% in F1 score and outperform the two baselines using dictionary or CRF separately.