Research On Text Categorization Technology Based On Deep Learning

Text categorization is one of the most common tasks in the field of natural language processing,the traditional text representation method is based on the word bag method,which is difficult to express complex semantics.In recent years,with the development of deep learning technology,the ability of capturing various text features is very strong,which has been paid much attention by researchers.Especially the LSTM model is put forward,which has achieved good results in some text categorization tasks.However,this paper holds that there are some deficiencies in these schemes,which do not fully excavate the potential of the deep learning technology in the task of text categorization.On the one hand,the current depth learning model captures the textual features when the angle single,and can not fuse multi-angle features;On the other hand,it is difficult for researchers to find a model that applies to many text categorization tasks,which need to be addressed for the task design model.In order to solve these problems,a text categorization framework based on model fusion for various text data and classification tasks has been proposed,which mainly includes the following research work:LSTM model as the cut-in point.(1)For the text basic unit usually uses the word vector,cannot completely contain the semantic information question,the character vector and the word vector synthesis processing mechanism has been proposed,which captures the text basic unit information from two kinds of angles.(2)In order to solve the problem of extracting text features from multiple angles,different models of syntax structure has been proposed,which can use order,reverse sequence and other angles to excavate semantics,and integrates them into a frame through the model fusion method.In the case of insufficient information of sentence-level text,Multi-angle semantic features are helpful to classify tasks.(3)Facing the problem that the multi angle feature dimension is too high and the classifier is difficult to deal with,a scheme of fusing multidimensional features with MLP layer in the feature processing part has been proposed,which can be used in most feature fusion tasks.In this paper,a text categorization framework based on depth learning technology has been presented,which has achieved good results in many experiments,and also proved the effect of some advanced learning techniques on text categorization task.It can provide the solution for some text categorization tasks,which has the dual value of theory and application.