The Modeling Research Of User Preference In Social Media

With the development of Web2.0,people have been used to publishing their opinions and views on the Internet,and obtaining needed information from others.This situation leads to a new Internet mode based on extensive users.In this Internet mode,people are dependent on Internet increasingly.In the beginning,Internet is a tool to search information,and then various communicate communities appear in the Internet,and now people would like to study others' reviews before they make a decision.The Internet is changing every aspect of people's daily life.The social media is just the medium for these behaviors.Social media is the virtual community or network platform where users can create,share and communicate their opinions,views and experiences,e.g.,microblog,blog,forum,online community,review site and so on.Users publish their opinions on the social media,and the personal opinions often contain abundant sentiments.These opinions can be divided into two types.One is the text information,such as messages in Twitter,while the other is the rating information,such as ratings for movies.User preference is the sentiment that a user shows about an event or an object,such as like and dislike.The research of user preference is to understand users' sentiments by researching the opinions which contain abundant sentiments.This paper studies the problem of user preference in social media by two sides:aspect rating prediction and sentiment analysis for the poems in Tang Dynasty.Aspect rating is the detailed rating on each aspect of the product,and overall rating is the comprehensive rating on all aspects of the product.Most of the existing works on aspect rating prediction have a basic assumption that the overall rating is the average score of aspect ratings or the overall rating is very close to aspect ratings.However,after analyzing real datasets,we find there is an obvious rating bias between overall rating and aspect ratings,and the existing works did not consider this rating bias.We are the first to study the problem of aspect rating prediction with rating bias,and design a novel sentiment-topic mixed model Rating-Center Model with Bias(RCMB).RCMB adopts the overall rating as the center of the probability model,and integrates the rating bias priori information with a latent aspect rating variable.Experiments on two real datasets(Dianping and TripAdvisor)validate that RCMB significantly improves the prediction accuracy over existing state-of-the-art methods,and maintains the relative order better.Existing researches of sentiment analysis focus on modem texts,such as product reviews and microblogging,and hardly involve the analysis of ancient literature.In this paper,we propose a model TL-PCO based on transfer learning to classify the poems in Tang Dynasty,and then we investigate social and cultural development in that era through analyzing the sentiments of ancient poetry.TL-PCO uses two proposed functions based on transfer learning to get two kinds of features.With the addition of features from ancient literature itself,three classifiers can be trained and then vote for the final category.Experiments demonstrate the effectiveness of the proposed method on the dataset of Chinese poems in Tang Dynasty.Moreover,the different periods of Tang Dynasty and different genres are analyzed in detail.Compared with the analysis of social history,the results confirm the effectiveness of this method.