Research On The Model And The Key Technologiesof Social Search Engine Based On HolonicMulti-Agent System

The development of Web 2.0 has made online social network applications popularized.A large number of users express opinions and share their lives on online social network platforms,so that a lot of user-generated content is producing.These data is useful for better search experience.In this context,social search engine,which aims to utilize social network data to optimize search results and improve user satisfaction with search engines came into being.However,there are still some problems in this area,including limited retrieval resource,insufficient knowledge application,no mechanism for active updating of social network knowledge,and fragmented research and no instructions for modeling designing.Therefore,this dissertation proposes a social search engine model based on the Holonic Multi-agent System which can retrieve resource covering all World Wide Web.This model utilizes the obtained multiple knowledge to optimize the search results from the component search engines,and personalized search result list can be provided.Moreover,the proposed model also has the ability to actively recommend diverse information that user might be interested in.The main contributions of this dissertation can be summarized as follows:(1)A model for HMAS based on the task-oriented perspective is proposed.By designing the holonic structure and modeling task related information,the static organizational structure model is studied.To adapt to changing environmental requirements,a dynamic self-adaptive mechanism which integrating the competitiveness adjustment and structure adjustment is discussed.Moreover,the competitiveness-based contract network protocol is utilized for the task assignment in a Holon,which can guarantee the success rate of systematic task.On this basis,this dissertation designs a social search engine model,which supports the adaptive task scheduling based on the proposed HMAS.The experimental results show that when the external environment changes,IM Search could trigger the competitiveness adjustment mechanism and the structure adjustment mechanism to adapt to the new environmental requirements.(2)A mechanism of data perception and knowledge updating for social search engine is designed.To realize the active perception of changing in social network data,the data perception rules are designed based on users' behaviors.These rules are continuously updated by learning users' behavior changes,ensuring the freshness of social network data.Moreover,to ensure the freshness of social network knowledge and reduce unnecessary system overhead,the knowledge updating method determines the changes in social network data could cause changes in social network knowledge based on threshold,and then update the related knowledge in a timely manner to achieve the active perception and updating of social network knowledge.The experimental results show that the proposed mechanism has an update rate of 92.6% for social network data changes,and the update rate of 72.5% for social network knowledge changes.(3)A personalized result ranking method based on query words recognition is proposed.The proposed method utilizes the influential members of the interest community to obtain the pseudo-related documents,and expands query terms to help users express search intention better.For non-navigational query words,this dissertation considers user interest drift,and utilizes multiple features to ranking documents,such as similarity between user interest topics and web document topics,the similarity between user interest keywords and web document keywords,and the characteristics of documents returned by component search engines,etc.The experimental results show that the proposed method performs better than the baselines.The NDCG value for the information queries is 0.673,while that of the transaction queries reaches 0.706.(4)A diversified information recommendation mechanism is proposed.According to the click records of user's community members,and the shortest path between the members and the user,the proposed mechanism recommends web documents that user might be interest in.Based on the influence knowledge and the social similarity between the current user and other users,the proposed mechanism could recommend other users in the query field that the current user might be interested in.This dissertation also provides hotspot information recommendation based on the click-through data of users' interest community and interactive community,the shortest path between the user and the community members is considered as well.The experimental results show that,for information queries,the MAP value of the proposed result recommendation method reaches 0.744,and that of transaction queries reaches 0.731.The proposed user recommendation method based on social network knowledge has a relevance score of 0.709,and the MAP value of the proposed hotspot recommendation method based on interest community and interactive community reaches0.750.The proposed information recommendation mechanism has the ability to recommend the required information for different users.This dissertation focuses on the existing problems of social search engine,and studies the overall design of social search engine model and key technologies such as algorithms,mechanisms and strategies.The proposed social search engine model for the World Wide Web,the active perception and updating mechanism of social network knowledge,the result ranking method and the information recommendation mechanism based on social network knowledge,could improve the precision ratio for social search engine.