Construction Of Linked Data Entity Type Prediction Model

Many linked data are incomplete in the type information of entities,the lack of which has challenged many Semantic Web-based tasks.The traditional type inference method can deduce the types of untyped entities through logical axioms,but this approach may be invalid in noisy data.The prediction results based on data-driven type prediction methods in recent years show that this method shields the data quality problems that may exist in linked data in different fields to some extent,and achieves the automatic completion of the entity type.However,the prediction effect of the existing data-driven entity type prediction method still needs to be further improved.Therefore,an effective data-driven entity type prediction problem remains to be studied.Based on the widely used collective classification algorithm ICA,this thesis uses entity type vectors constructed by extracting linked information with different indications for entity type prediction to perform entity type prediction.In this process,in view of the time efficiency problem of the original ICA applied to the large linked datasets and the correct rate of the entity type prediction result when the entity type prediction is performed,this thesis proposes a method by setting the condition for updating entity type vector,and a method of entity type prediction by constructing an entity type vector having a stronger positive influence on an entity type prediction result by utilizing richer linked information.Among them,different from the linked information used by the original ICA to construct the entity type vectors,the information used to construct the entity type vectors in this thesis includes three categories,they are data attribute information,neighbor entity type information and object attribute information.Finally experiments on multiple real linked datasets show that the method of entity type prediction based on ICA in this thesis very effective in finding missing types,and provides new ideas and methods for entity type prediction by making full use of relevant information in the A-box of linked data,at the same time,it has certain reference value for the automatic completion of entity type based on linked data.