Chinese News Text Classification Method via Key Feature Enhancement

(1) Background: Chinese news text is a popular form of media communication, which can be seen everywhere in China. classification an important direction natural language processing (NLP). How to use high-quality technology help humans efficiently organize and manage the massive amount web urgent problem solved. It noted that existing deep learning methods rely on large-scale tagged corpus for tasks this model poorly interpretable because size large. (2) Methods: To solve above problems, paper proposes method based key feature enhancement named KFE-CNN. effectively expand semantic information features enhance sample data then combine zero–one binary vector representation transform into vectors input them CNN training implementation, thus improving interpretability compressing model. (3) Results: The experimental results show our significantly improve overall performance average accuracy F1-score THUCNews subset public dataset reached 97.84% 98%. (4) Conclusions: fully proved effectiveness KFE-CNN task it also demonstrates performance.