Research Of Fine-grained Sentiment Analysis For Product Reviews

With the rapid development of e-commerce platforms,more and more people are keen on online consumption,followed by massive product reviews.And these product reviews often contain a lot of information,carrying the user's emotional tendencies towards the various attributes of the product.In the fine-grained sentiment analysis of products,the evaluation object and evaluation word extraction are the most critical.Only the accurate extraction of evaluation objects and evaluation words can accurately analyze the user's emotional tendency towards each attribute of the product.In this thesis,a new model is proposed to improve the extraction accuracy of evaluation obj ects and evaluation words.Understanding whether a user's attitude toward a particular attribute of a product is like or dissatisfied,a fine-grained sentiment analysis can achieve this.In the case of fine-grained sentiment analysis,our first step is to find the evaluation object and evaluation word in the product review text and match them one by one.Based on this,this thesis proposes a new algorithm GLDA,which is composed of a deep neural network gated recurrent unit(GRU)and a topic model potential Latent Dirichlet Allocation(LDA).In order to further improve the extraction accuracy of the evaluation object,this thesis constructs a collection of evaluation objects for specific fields,and realizes clustering of domain synonyms(different evaluation words used to describe the same feature is called domain synonym),which improves the extraction accuracy of the evaluation object.Based on the proposed new algorithm GLDA,this thesis designs a model suitable for fine-grained sentiment analysis of hotel reviews,and applies this model to the real review data of the hotel on the TripAdvisor website,and makes a fine-grained sentiment analysis on the comment data.The experimental results show that the proposed new algorithm GLDA is better than LDA and SentenceLDA(Sentence Latent Dirichlet Allocation)in text modeling ability,and the analysis of the fine-grained sentiment analysis model based on GLDA algorithm on hotel data is better than the other two.Algorithm accuracy and recall rate are high.The availability of the model was confirmed.