News-Comment Relevance Classification Algorithm Based on Feature Extraction
Hongxia Wei, Wenguang Zheng, Yingyuan Xiao, Chen Dong · 2021
In recent years, with the rapid development of the mobile Internet, it has become easier for users to read news and corresponding comments. Most people get used to reading news on-line. However, sociologists have shown that up to 50 percent of the comments in a news article are irrelevant to the news content. In this paper, we investigate the news-comments relevance analysis problem. We formalize such problem as a classification problem and proposed a news-comment relevance classification algorithm, based on BERT feature extraction. This algorithm uses a model trained in the extensive semantic database to extract the feature vectors of news and its comments and inputs the extracted feature vectors into a full-connection layer network to complete the classification task. Several experiments are carried out on five public datasets, and the experimental results show that the model I proposed has good performance.