Intent-aware Personalized Content Search Across Multiple Social Networks

With the rapid development of the Internet,social networks such as Twitter,Facebook,Instagram,Sina Microblog,Zhihu,Douban etc.are proliferated.Online social network services are becoming more and more abundant,and online users are gradually increasing.Users frequently communicate and interact with each other through various social networks,which leads to the rapid expansion of information in online social networks.Meanwhile users are used to retrieving information across social networks.Traditional search engines only focus on the speed of finding results and the popularity of links,but lack of analysis of users' search intention.In addition,more and more users use more than two social networks,and the existing social search works are mainly based on a single social network.A few works of cross-social network information retrieval not only ignore the information barriers between different social networks,but also ignore the impact of different social networks on users.In view of the weakness of previous researches,a novel intent-aware personalized content search across multiple social networks model is proposed to analyze multi-modal data on multiple social networks.The model analyze users' interest preferences and platform preferences,perceive users' current search intentions,and provide personalized content search across multiple social networks.The main work includes:Firstly,this thesis select seed users from Aboutme,and capture their followees and their tweets on multiple social networks including Twitter,Instagram.On this basis,text preprocessing is carried out to reduce the irregularity of text contents.In addition,in order to break the information barrier of Twitter text and Instagram images and solve the problem of multi-modal data analysis,it is also necessary to preprocess the captured image content.Furthermore,a search model based on intention perception is proposed.This model takes into account the feature that social data are generated independently by users,and uses the multiple social networks topic model to model users' references,perceive users' personalized search intentions,and build a bridge between multi-modal data.At the same time,this model designs the online and offline search ranking algorithm to provide multiple social networks information for users.On this basis,the topic model variable sampling and parameter updating rules designed in the model are deduced and demonstrated,and a sampling algorithm for multiple social networks topic model parameters estimation is proposed.Finally,experiments are carried out based on a real dataset of multiple social networks.Through the comparison and analysis of the experimental results,we can get the following conclusions: the model can effectively model users' preferences,perceive the users' search intention,and fuse different social network data.Using multi-modal data to supplement user information can effectively break the barrier of different social networks information,and provide users with more personalized social search content.At last,an intent-aware personalized content search across multiple social networks system is designed and implemented.