Research On Information Recommendation Based On Entity Linking In Dual-structural Network

With the advent of the big data age,the information on the Internet is growing exponentially,and in face of massive amounts of information on the Internet,it is becoming increasingly difficult for people to find useful information for their limited needs quickly and get rid of spam.The dual-structural network argues integrating the current Internet architecture with a broadcast-storage secondary structural network characterized by "radiation-copy" model,to realize the "de-redundancy" of the current Internet with the innovative ideas of physical transformation and dual structure,and providing distinctive personalized services in the user terminals to help users quickly select useful information.However,the traditional recommendation algorithm is too dependent on user interaction behaviors so that it can not dispense the cold start problem.The ontology-based semantic recommendation algorithm is immune to the above problems,but suffers poor real-time performance.Meanwhile the dual-structural network generates tens of thousands of new information every day,the timeliness requirement is high.How to minimize the dependence on user interaction and the impact of the cold start problem,and quickly produce accurate recommendation result is a problem of the recommendation mechanism in dual-structural network.For the problem that the traditional recommendation algorithm relies too much on the user interaction behaviors and dissatisfies the timeliness requirement of the dual-structural network,this thesis uses Named Entity(NE)to represent the semantic meaning of information,proposes a Dynamic Semantic Associative Entity Library(DSAEL)and an Entity Linking Method based on Dynamic Semantic Associative Entity Library(DSAEL-EL).Besides,depending on DSAEL-EL and the characteristics of dual-structural network,this thesis raises a Hybrid Information Recommendation Algorithm Combining Entity Semantics and User Entity Interesting Tag(HRESIL),to help users generate real-time information recommendation.The main work of this thesis is as follows:1)Focusing on the problem that traditional physical linking method does not apply to dual-structural network,this thesis designs an entity linking method DSAEL-EL based on DSAEL.First,this thesis constructs the DSAEL base library by using Wikipedia and Baidu Encyclopedia dumps corpus,then identifies the entities in the information and calculates their semantic weight,and then determines the target entity by calculating the similarity between the associative entity set of the candidate entity and the undisambiguated entity set of the information,and finally uses the target entity set to update the DSAEL.DSAEL-EL performs entity disambiguation according to the semantic similarity of entity,which allows it to produce more reasonable linking results adapting to changes in context.2)Aiming at the problems that the traditional recommendation algorithms rely too much on user interaction behaviors and perform poor on timeliness,this thesis puts forward a hybrid information recommendation algorithm HRESIL which combines entity semantics and user entity interest labels.At the broadcast source point,this thesis obtains the target entity set of the information by the DSAEL-EL algorithm,and updates DSAEL with the target entity set.Then,this thesis indexes the target entity set into Uniform Content Label(UCL)and distributes UCL to the edge servers and the user terminals.The edge servers update the DSAEL sub-library with the received UCL,and interact with the user terminals to response the users’ personalization requests.Compared with the traditional recommendation algorithm,HRESIL does not depend on the users’ interactive behaviors,and generates the recommendation information list through the DSAEL node association,which makes it extremely suitable for the dual-structural network.3)Based on the dual-structural network prototype system,this thesis designs and implements a dual-structural network news recommendation system with DSAEL-EL and HRESIL algorithms,and verifies their performance by related experiments.The experimental results show that DSAEL-EL owns higher accuracy than the traditional entity linking algorithm.Compared with the traditional recommendation algorithm,HRESIL suffers much less from the cold start problem and has better timeliness.