Research On The Personalized Recommendation Technology Based On Sentiment Analysis

At present,the Internet has become a part of people's work and life,users can communicate,express their personal opinions and comment on various products on the network.The amount of information online is growing exponentially,causing information overload.Personalized recommendation technology provides users with a powerful means of information filtering and provides users with interested information.Traditional recommendation methods often make use of users' historical behaviors.With the increase of comment information,how to make use of user comment information to improve the efficiency of personalized recommendation has become a research hotspot.Based on user comment information,this paper integrates textual comment sentiment analysis technology into personalized recommendation system,and proposes a personalized recommendation model based on aspect category sentiment analysis to improve the accuracy of recommendation.The main research contents include:(1)Data analysis and preprocessing: conducted in-depth research on aspect-based sentiment analysis technology and personalized recommendation technology,and preprocessed Semeval-2014 data set and Yelp data set,including analysis,data cleaning,case conversion,word segmentation,part of speech restoration,stem extraction,destop words,text vectorization,etc.(2)Aspect-based Sentiment Analysis Model: This paper studies the application of bidirectional short and long time memory network and aspect embedding gated convolutional network in sentiment analysis,and compares their advantages and disadvantages.An aspect category sentiment analysis method of BiLSTM?GCAE which combines the two is proposed,and a comparative experiment is carried out.,which is verified by comparative experiment based on Semeval-2014 data set,the accuracy rate is 80.20%.Experiments show that the performance of this model is improved in sentiment classification,and it is better than the traditional sentiment analysis method.(3)Neural Network Collaborative Filtering Recommendation Model Based on Sentiment Analysis: Aiming at the problems of information loss and insufficient interpretability in the current recommendation system,a neural network collaborative filtering model based on aspect sentiment analysis is proposed by combining Sentiment Analysis and Neural Network Collaborative Filtering.First,the text reviews for fine-grained analysis of the emotion,with emotion said user preferences and the characteristics of the commodity said model,then the user and the characteristics of the commodity said after a bitwise by operating the vector into the neural network learning to get higher order implied characteristics,improve the ability of nonlinear processing,so as to realize more accurate recommendations,A comparative experiment was conducted using Yelp dataset.