Research And Implementation Of Text Sentiment Analysis System Based On Neural Network Model

Since the beginning of the 21st century,the development of deep learning framework has made breakthroughs in many field of image and speech.In the field of natural language processing,the deep learning models have achieved good results on many traditional problems,especially in the text classification task where some typical neural network application improved the accuracy of the text classification to a great extent.In the whole,there are two steps that are critical when using the neural network models based on the deep learning neural network:1 Converting the words in the text into word vectors;2 Encoding word vectors sequence as sentence vectors.The existing deep learning models in the text classification have achieved good results,but in these two steps there are also some shortcomings.In this paper,the following two improved network structures are proposed for the defects of existing models in these two steps.Firstly,in the process of converting words into word vectors,the existing models based on deep learning directly used unregulated word vector generation strategy such as word2vec.Words are regarded as a basic and indivisible unit in languages.They are mapped to vector space through the probability of coherence between words.These methods just consider the statistical position relationship between the words,and do not take the structure information in the words into account.Whether in Chinese or English,the generation of words have a certain regularity.This paper put forward the neural network text sentiment analysis model based on word-building method to automatically extracted the structural information from the words through a neural network structure.Secondly,in the process of encoding the word vector sequences into the sentence vectors,the existing models based on deep learning mostly use the convolutional neural network or the recurrent neural network.However,the convolutional neural network has local flaws in the extraction of sentimental characteristics and the recurrent neural network has the defects of sequence bias in extracting the sentimental features,which would affect the classification performance of the final model.In this paper,we put forward a text sentimental analysis network based on sentimental information collector-extractor architecture and used a reasonable way to combine the recurrent neural network with the convolutional neural network.Based on the above two kinds of improved neural network structure,this paper studies and implements the weibo text sentiment analysis system.This paper introduces the four main functional entities of the whole system:1.data fetching and preprocessing module;2.data analysis module;3.data storage module;4.data display module.At the same time,the inner design details of these four functional entities are introduced in detail,and the system test and front-end display are carried out in the end.