UCL Knowledge Space Based On Entity In Dual-structural Network

Dual-structural network(DSN)is a new type of network architecture with Internet as its primary structure and broadcast-storage network as its secondary structure.It can fully utilize physical broadcast to distribute pop content to users,and can effectively reduce the number of routing hops for content distribution.The Uniform Content Label(UCL)is used to solve the shortcomings of the lack of interactivity in asynchronous broadcast communications,and helps to straighten out the unstructured content for Internet,enabling users to obtain rich semantic information in Web content fleetly.However,the current research on DSN focuses on the caching strategies and information recommendation so that UCLs that can be acquired by users being limited to users' interests and local caches,and users are lack of awareness of global UCL information.Therefore,it is of great prospective significance to study how to effectively organize the scattered UCLs semantically to provide a reliable UCL knowledge discovery service for users.For this reason,this thesis proposes a UCL semantic association model based on Named Entity(NE),and designs the UCL Knowledge Space(UCLKS)to realize the knowledge representation,organization and discovery of massive UCLs.On the basis of UCLKS,a solution for UCL content analysis service for natural language questions is proposed so that users can obtain global UCL knowledge easily.Specifically,there are three major contributions in this thesis as follows:1)In order to effectively organize the massive UCLs in the DSN and combine the requirements of UCL knowledge discovery,the thesis designs a knowledge space based on UCL,and proposes a construction method of UCLKS based on Baidu Encyclopedia.Firstly,this method builds an association graph based on named entities by crawling Baidu Encyclopedia entries.Then,the entities in UCL are extracted using a combination of machine learning and rule-based method and the semantic weights are calculated.The UCL and the named entity association graph are merged by means of an entity disambiguation algorithm.Finally,a persistent solution for UCLKS is proposed.The UCLKS semantically fuses encyclopedia knowledge with UCL information through named entities,providing a reliable basis for UCL knowledge discovery.2)Aiming at the problem of UCL knowledge discovery in DSN,this thesis proposes a UCL-based content analysis service and designs a solution of UCL content analysis service that can understand natural language questions.Firstly,the solution translates the natural language question into a database query statement through a query generating algorithm based on a word-order deduction rule.Then,the relevant UCLs are acquired in UCLKS according to the users' requests.Finally,the UCLs are sorted according to the degree of semantic association with the users' questions and returned to users.The solution of UCL content analysis service can understand users' questions and provide users with encyclopedia knowledge and related UCL information.3)On the one hand,a content analysis prototype system based on UCLKS was developed which based on the DSN prototype system of our research group.The prototype system semantically relates UCLs through named entities to establish the UCLKS,and answers users' content analysis requests through the semantic understanding of users' questions.On the other hand,the entity disambiguation algorithm based on context similarity,query generation algorithm based on a word-order deduction rule,and UCL sorting algorithm based on semantic association are validated and analyzed by using real content of the prototype system.Experimental results verify the feasibility and effectiveness of the proposed algorithms.