Research On Chinese Text Classification Based On Convolutional Neural Networks

In today’s era of explosive internet development,the internet has led to an increasingly wide range of channels for news dissemination,as well as an increasing number of news texts,resulting in users lacking the energy to search for specific news categories from massive amounts of news.In the face of text classification in the field of natural language processing,the neural network method,which has obvious advantages in the screening and prediction of big data,can play an important role.Textcnn is a text classification algorithm based on convolutional neural networks.This article proposes an improved method to address the weak feature extraction and classification ability of the Textcnn model in text classification.To improve the ability of model feature extraction,this paper proposes a text feature extraction algorithm with fusion factors.This article improves the traditional TF-IDF algorithm by incorporating part of speech and position information to adjust the feature weights of words in text,and proposes an LTF-IDF＿POS algorithm;On this basis,this article proposes a segmentation feature extraction algorithm based on sliding window technology.Integrating the above two parts,this article proposes an improved feature extraction algorithm that improves the problem of unlisted words in word segmentation algorithms,as well as the shortcomings of existing feature extraction algorithms that do not utilize the distribution information of words within the text.After conducting experiments on this algorithm,the results show that it performs better in news text feature extraction compared to other traditional algorithms and helps improve the ability of text classification.To solve the problem of insufficient classification ability of the model,this article combines the Word2 vec word vector model with the Textcnn model.The Word2 vec model utilizes the contextual information of words in the text to transform them into a low dimensional vector,which can then transform an article into a dense matrix.For convolutional neural network model,this paper modifies its activation function to improve the classification efficiency of Textcnn model.In order to verify the effectiveness of the improved model,the model was tested on the dataset.The test results showed that the accuracy of the improved model was improved compared to other models,and it can better improve the performance of text classification.