Original scientific paper

https://doi.org/10.17559/TV-20190623122323

News Text Classification Based on an Improved Convolutional Neural Network

Wenjing Tao ; Beijing Jiaotong University, No. 3, Shangyuan Village, Haidian District, Beijing, China
Dan Chang ; Beijing Jiaotong University, No. 3, Shangyuan Village, Haidian District, Beijing, China

Abstract

With the explosive growth in Internet news media and the disorganized status of news texts, this paper puts forward an automatic classification model for news based on a Convolutional Neural Network (CNN). In the model, Word2vec is firstly merged with Latent Dirichlet Allocation (LDA) to generate an effective text feature representation. Then when an attention mechanism is combined with the proposed model, higher attention probability values are given to key features to achieve an accurate judgment. The results show that the precision rate, the recall rate and the F1 value of the model in this paper reach 96.4%, 95.9% and 96.2% respectively, which indicates that the improved CNN, through a unique framework, can extract deep semantic features of the text and provide a strong support for establishing an efficient and accurate news text classification model.

Keywords

attention mechanism; Convolutional Neural Network (CNN); feature representation; text classification; Word2vec

Hrčak ID:

226037

URI

https://hrcak.srce.hr/226037

Publication date:

8.10.2019.

Visits: 2.305 *