BERT based Web Mining of Concerns and Reviews for TV Drama Audience

This paper proposes how to mine concerns and reviews that are relevant to TV dramas from the results of collecting Web pages on those TV dramas. The proposed framework consists of the natural language processing based techniques of analyzing sentence embeddings of those collected Web pages. More specifically, we apply the technique of BERT to the task of judging whether the theme of a collected Web page is relevant to the given drama. We also apply the technique of BERT to the task of judging whether a collected Web page includes any review or not. Results of evaluating those proposed models show that the proposed framework performs well in those tasks of relevance judgment and review detection.