| In the era of big data,there are more and more applications.In order to improve user experience and increase user stickiness,Personalized Recommendation Algorithm has become the key field of competition among major applications.At present,Personalized Recommendation Algorithm mostly studies users’ preferences based on their previous product ratings,and then provides them with recommendation list.However,the network user’s rating is too concentrated,and the rating credibility is not high,resulting in the recomme ndation accuracy is not ideal.As the network information becomes more and more abundant,there are more and more information that can reflect users’ emotions.In addition to users’ ratings,users’ reviews can also be used as the basis of recommendation al gorithm.Therefore,Personalized Recommendation Algorithm based on emotion analysis stands out from all kinds of recommendation algorithms and becomes a hot topic in current research.In order to screen out the user reviews that express the user’s feelings,Chinese Word Segmentation technology is needed.In order to turn these emotional information into value,we need to design a Quantitative Model of Emotions.In order to achieve accurate recommendation,Collaborative Filtering Recommendation Algorithm is needed.This paper focuses on these issues: how to combine Chinese Word Segmentation technology and Emotion Analysis technology into Collaborative Filtering Recommendation Algorithm to realize Personalized Recommendation Algorithm based on emotion analysis and improve the recommendation accuracy.To solve the problem that low credibility and low differentiation of movie ratings affect the effect of Recommendation Algorithm,an improved Collaborative Filtering Algorithm is designed in this paper: We use Chinese Word Segmentation technology and emotion analysis technology to separate the emotion words,negative words and adverbs of degree.The quantitative value of the emotional attitude calculated by the model is used to modify the score so that it is closer to the user’s real score.The modified score was substituted into the Collaborative Filtering Recommendation Algorithm to produce more accurate recommendation results.This paper quantifies the emotion of users’ movie reviews by using every emotion word,negative word and degree adverb in the reviews.Each positive emotion word represents unit 1,and each negative emotion word represents negative unit 1;Each negative word changes the emotional polarity of the first emotion word that comes after it;Each adverb of degree expands or reduces the weight of the emotional intensity of the first emotion word that appears later.In this paper,we transform the problem of combining comment and score into a single numerical optimization problem.Assuming the modified score y=w*score+(1-w)*degree,the w with the highest accuracy and recall rate of the recommendation algorithm is selected as the parameter through experiments.According to the particularity of the data in this paper,the user-based Collaborative Filtering Algorithm and the item-based Collaborative Recommendation Filtering Recommendation algorithm are respectively used in this paper to select the one with better recommendation effect by comparing the experimental results.This paper has done a lot of experiments and studies.Finally,the experimental results show that the improved recommendation algorithm based on emotion analysis has been improved in both accuracy and recall rate.In other words,the recommendation algorithm based on comprehensive consideration of film review and score is better than the recommendation algorithm based on only film review or score. |
