News Text Classification Based on Deep Learning and TRBert Model

The utilization of the Internet and various social media platforms has led to an increase in the amount of text information available to people. Natural language processing technology has become increasingly practical for dealing with large amounts of text data, with news text classification being a particularly important task in terms of formulating news retrieval strategies, news recommendations, and social public opinion monitoring. While news texts on the Internet are typically divided into titles and texts, most existing deep learning-based research does not take into account the processing of titles and texts. The single processing method of merging titles into texts often results in unsatisfactory classification results. This chapter will focus on the importance of headlines in news texts, which are summaries of the main content of the news, contain a great deal of information, and express the central idea of the news. The TRBert model is proposed to complete text classification tasks, which incorporates the attention mechanism and hierarchical BERT weighted title word vectors.