| With the rapid development of science and technology,Text information is growing rapidly.How to deal with huge and valuable text information and extract valuable information from it has become the focus of modern scientific research.Text classification is one of the important research directions.This paper studies the text classification method,and based on word vector to improve text classification precision.First,aiming to solve the problem of high feature dimension and semantic ambiguity,the word vector is used to represent the vocabulary and the text.Word vectors have low dimensionality and make full use of the context information of the text,so that word vectors have rich semantic information.Second,a text classification method of mutual information weighted word vector is proposed which combines text statistical information and semantic information.Mutual information reflects the statistical features,and word vectors reflect the semantic features.Weighting the word vector of text feature words with mutual information and summing them as text vectors.Text features are extracted from multiple dimensions.Then text classification precision is improved by using support vector machine(SVM)and decision tree(DT)algorithms.Next,text is the characteristics of serialized data,using the recurrent neural network(RNN)structure to automatically extract text features.A weighted recurrent neural network(W-RNN)is proposed based on RNN.The automatic weighted summation of the intermediate output of each word vector forms a text representation vector,classification using a simple neural network model.This method can fully extract text serialization semantic information and high-level information and obtain a better result.Finally,the text classification methods are experimented on the news dataset to validate the effectiveness of the proposed methods. |
