Sentiment classification method of hierarchical retrieval based on comment data

In order to help enterprises and the news industry liberate human resources, quickly understand the public opinion guidance of news reviews, and improve the quality of product content, this paper, aiming at multi-source digital news reviews, combined with crawler technology and Bert pre training model, carries out experiments on DNN, CNN, LSTM and other deep learning models. According to the particularity of the samples, a hierarchical retrieval emotion classification method is designed and proposed, which is applied to the field of news.