基于加权Word2Vec和TextCNN的新闻文本分类

News text classification based on weighted Word2Vec and TextCNN

随着网络和各类社交媒体的盛行,越来越多的文本信息通过互联网呈现在人们面前。对于海量的文本数据,自然语言处理技术变得越来越实用,新闻文本分类便是其中一项重要的任务,其对制定新闻检索策略、新闻推荐、社会舆情监控等具有积极作用。文章通过分析文本表示模型与分类模型的研究现状,提出一种基于加权Word2Vec和TextCNN的新闻文本分类方法,在新闻文本多分类数据上进行实验。从实验结果上来看,在文本表示模型中,该文方法比TF-IDF模型、Word2Vec模型以及随机词嵌入模型在精确率、召回率和F1值上均有提高;在文本分类模型中,文章使用的TextCNN模型要比传统的机器学习模型以及循环神经网络模型在分类效果以及模型性能方面表现更出色。

With the prevalence of the Internet and various types of social media, more and more text information through the Internet presented to people. For massive text data, natural language processing technology has become more and more practical, news text classification is one of the crucial tasks, which has a positive effect on the development of news retrieval strategies, news recommendations, social public opinion monitoring, etc. By analyzing the research status of text representation model and classification model, this paper proposed a news text classification method based on weighted Word2Vec and TextCNN, and experiments were performed on multi-classification data of news text. From the experimental results, in the text representation model, the proposed method is higher than the TF-IDF model, the Word2Vec model and the random word embedding model in terms of accuracy,recall rate and F1 value. In the text classification model, the Text CNN model used in this paper performs better in terms of classification effect and model performance than the traditional machine learning model and the recurrent neural network model.