Sentiment Classification And Opinion Extraction Of Online Reviews

Sentiment analysis refers to the use of text mining and natural language processing technology to analyze and process subjective comments with emotional.Sentiment analysis includes Sentiment Classification,Opinion Extraction and Opinion Question Answering.This paper studies Sentiment Classification and Opinion Extraction of online reviews.In the sentiment classification part,this paper improves on the basis of the Bi-LSTM model.The specific improvement methods are: introducing Convolutional Neural Network and Attention Model.The reasons for the improvement are as follows: firstly,the Convolutional Neural Network is introduced to extract the local features of the text,in order to improve the feature extraction ability of Bi-LSTM;secondly,the Attention Model is introduced to give a higher weight to the key information in the text,which improves the problem that Bi-LSTM cannot highlight important information.The experimental results show that the accuracy、precision、F1 value and recall of the model are improved,so the model improvement is effective.In the opinion extraction part,this paper constructs the BERT+CRF model to extract attribute feature words and opinion words in the comments;attribute feature words are generally words that describe a feature of a product or event,and opinion words are used to express user’s attitude.Then,this paper build the BERT model for Opinion Classification and attribute feature classification,get the categories corresponding to the attribute feature words and the sentiment polarity of the opinion words,and combine the extraction results to obtain a fine-grained evaluation quadruple: {Aspect Terms、Opinion Terms、Categories、Polarities}.Experiment shows that the model constructed in this paper can extract opinions from online reviews well.