Research And Application Of Relational Knowledge Discovery Technology Based On Entity Linking

Today's world is in a period of information explosion,it is difficult and necessary to retrieve the needed information from the complicated information.As a result,knowledge discovery technology has developed rapidly,and the associated knowledge discovery system based on knowledge graph can mine deep related information by using entity and relational information.However,because the network information has the characteristics of mixed noise and sparse knowledge,how to transform it into structured knowledge in knowledge graph is a hot topic at present.Through the cleaning of the natural text,the lack of knowledge is screened out,and finally the entity link technology is used to construct the relationship between the natural text and the knowledge graph structured knowledge.At present,most entity link models lack the processing of text noise and the accurate expression of semantic features,and ignore the rich information of the knowledge graph itself,which leads to the poor effect of entity linking.For this reason,this thesis proposes a local entity link model based on attention mechanism and a collective entity link model based on information fusion.The main work and contributions of this thesis are as follows.In view of the insufficient semantic interaction between the input text and the knowledge base entity description in the traditional model,the semantic loss is serious,and the model can not learn the semantic representation completely.In this thesis,a local entity link model is proposed.By extracting the entity to mention the internal semantic features and interacting with the features described by the entity,the noise is reduced and the keywords are highlighted.Combined with the self-attention characteristics at the learning collective level,the problem of semantic information loss can be solved.The experimental results show that,the effect of the local entity link model proposed in this thesis is improved,and the effectiveness of the model is verified.In order to solve the problem that the traditional model ignores the characteristic information between knowledge graph entities and can not effectively establish the relationship between entity nodes and realize the "flow" of information between nodes,this thesis constructs a collective entity link model of multi-information fusion.analyze and extract rich features among candidate entities and fully depict the relationships between candidate entities.The experimental results show that,compared with the baseline model,the effect of the collective entity link model proposed in this thesis is improved,and the effectiveness of the model is verified.Finally,based on the above model,the knowledge discovery system in the film field is constructed,which realizes the basic functions such as user query and knowledge management,meets the requirements of the knowledge discovery system,and verifies the availability of the algorithm.The construction of the knowledge discovery system can find the hidden information related to the film,which has a strong application value.