Opinion Mining And Sentiment Analysis Based On Deep Learning

Nowadays,data,as a new generation of economic engine,has attracted more and more researchers' attention.With the development of the Internet,people communicate,comment,shop and express their opinions through the Internet which becomes increasingly pervasive.And an overwhelming number of text data accumulated with a high speed.However,how to get emotional information accurately from text data is still a research hotspot but also a difficulty in data science,intelligent computing and other fields and is expected to be mined.According to the point of network text,this paper mines and analyzes the view data and emotion discrimination of network text from the perspective of data science.This paper analyzes the problems existing in the network about text view mining:(1)the granularity of text view mining is single,the granularity is rough,the data is sparse,the data is unbalanced and so on.(2)Traditional mining needs a certain amount of manual processing,what's more,has the disadvantages of relatively simple features,low processing efficiency.To address these problems,deep learning method was adopted to aim at the research and improvement of the network text view and emotional data.Deep learning has been applied in many fields.Deep learning has also achieved excellent results in machine translation and text dialogue,which proves that deep learning has unique advantages in text processing.In this paper,neural networks with sequential mechanism,such as RNN,LSTM,GRU and so on,are used to extract and mine views and analyze emotional attitudes.Based on the above problems and research,this paper has done the following work:(1)From the perspective of data science,this paper illustrates the processing flow of the Network Text View,and adopts the deep learning method to integrate the attention mechanism into the deep learning model analysis of the sequential network,that is,the network text view information represented by combining the attention mechanism with the bidirectional recurrent neural network to train the model with emotional classification.The validity of the model is verified by experiments.(2)Combined with the characteristics of neural network,the bidirectional recurrent neural network is combined with the strong feature representation of the capsule network.Thus,the network model with complementary advantages is realized,the bidirectional recurrent neural network strengthens the network effect and the expression efficiency of viewpoint text.Attention selectively removes redundancies and reinforces useful ideas.The capsule network further strengthens the ability of classification and discrimination in this field.