Research On Text Classification Method Based On Convolutional Neural Network

With the advent of the mass information age marked by the Internet,Big Data and Deep Learning,how to excavate a large amount of useful information from these text data has become a hot research topic,which has very important research and application value.Therefore,text mining technology for the text information has been widely concerned.Text classification is a core research task of text mining,which has attracted much attention from academia and industry.Since the concept of deep learning was first put forward in 2006,deep learning technology has made great breakthroughs in the fields of image recognition,speech recognition and Machine Translation,which greatly promoted the development of artificial intelligence.Compared with traditional machine learning algorithms,deep learning technology has better performance.Convolutional neural network,as a typical representative of deep learning technology,has been proved to be very effective in image classification and image recognition field.This paper tries to explore the use of convolutional neural network to extract text features.Therefore,the following works are mainly done in this paper.(1)this paper mainly studies several common text representation learning methods in the text classification task.First of all,we introduced several language models,such as neural probabilistic language model by Bengio et al.(Neural Probability Language Model),C&W model by Collobert and Weston,BOW and Skip-Gram model by Mikolov et al.,and compared several models in terms of performance in two tasks.Word Embedding is considered as a byproduct of language models.Through several comparative experiments,we found that vectors generated by Skip-Gram model are more suitable for hierarchical convolution neural network based text classification model that was designed in this paper.(2)In this papr,a hierarchical convolution neural network model is designed to complete the text classification task.Text data in classification tasks often have hierarchical relations,for example,a sentence is composed of many words,a paragraph is composed of many sentences,and an document is composed of multiple paragraphs.This hierarchical relationship often exists in many text data.Therefore,in order to better extract this hierarchical feature,we design a hierarchical structure based convolution neural network classification model,which can extract this hierarchical feature well,thus enriching the feature space.Compared with traditional convolutional neural network based classification models and LSTM network based classification models,the convolutional neural network classification model designed in this paper has better performance and better classification effect.