Research And Implementation Of Personalized Search Engine For Social Network

In recent years,search engines have become the main tool for Internet users to acquire knowledge.With the development of social networks,traditional search engine that only provides a single search service can not meet the requirements of the users in various social networking sites.The hope that the search engine can return personalized search results according to different users,has become an urgent need for Internet users.However,the ex-isting personalization of search engines has mainly focused on the analysis of users' search history and click history,and has not used various features of users such as social relation-s and social interest preferences in social networks.Therefore,it is impossible to provide users with the individualized search results taking the user' multi-faceted needs into account.In this paper,the real data of the Zhihu social network is obtained through the third-party API interface.Firstly,the paper analyzes and screens the user data characteristics of the network,and then establish the user model for Zhihu network.The model is mainly composed of user topic model,social behavior model and influence model.Besides,the calculation method and updating mechanism of each model are designed respectively.In the establishment of the influence model,this paper proposes the Up Vote_Rank algorithm,which based on the various attributes of the network users.The algorithm is mainly im-proved based on the PageRank algorithm,not only use the topology attributes of the user in the social network,but also use the social performance of the user in the Zhihu network,including the number of Agree and Thank of the user.So that the final measure of user influence,can really reflect the importance of the user in the Zhihu network.Based on the Zhihu user model established in this paper,the paper designed a personalized search strategy,including the evaluation and sorting mechanism of personalized retrieval re-sults.In particular,based on the traditional BM25 weight model,the retrieval results' score are personalized by combining the user influence,the social behavior similarity between users and the topic model similarity between the user and the retrieval result,then sort the search results based on the score.The experimental part used the real data to experiment the strategy,and validated the effectiveness of the three sub-models respectively,including user influence model,user topic model and user social behavior model.Finally merge the three models,and then validated the validity of the combined effect of the user model.Finally,this paper designed and realized the personalized search engine for the Zhihu net-work.Zhihu users can use the system to retrieve questions which they interest.The system can take into account the user's various attributes,and provide users with personalized search results,making the results ranking forward can meet the user's social preferences.