Research on Aspect-Level Sentiment Analysis Based on Text Comments

Sentiment analysis is the processing of textual data and giving positive or negative opinions to sentences. In ABSA dataset, most sentences contain one aspect sentiment polarity, have multiple identical polarities, which weakens polarity dataset. Therefore, this paper uses SemEval 14 Restaurant Review in each document symmetrically divided into individual sentences, two versions datasets ATSA ACSA are created. ATSA: Aspect Term Analysis Dataset. ACSA: Category order simulate complex relationship between contexts accurately extract emotional features, combines latest development trend NLP, capsule network BRET, proposes baseline model CapsNet-BERT. The experimental results verify effectiveness model.