The Construction Of The Scholar Academic Network For Internet Open Resources

The construction of academic network based on open Internet resources can realize the multidimensional description and presentation of academic information,and help researchers to identify the needed academic resources.What’s more,there are some problems in the construction of academic network,such as the difficulty of information entity extraction and the dynamic evolution of academic network.This paper investigates the dynamic construction of the scholar academic knowledge network for open resources based on the named entity recognition,relation extraction,and entity alignment.Then scholar information entity recognition and the method of the scholar academic knowledge network dynamic evolution are proposed.Using these methods,the scholar academic knowledge network achieves the automated integration of scholar information and the evolutionary update.It can provide the basis for the application of scholarly data.The contributions consist of three sections.(1)Scholar information entity recognition for open resource.Scholar text information obtained from the Internet lacks context information related to the entities.This paper feeds these texts into Google search engine,and uses BERTScore to caculate the similarity of the search query and retrieved texts.Based on similarity,it can construct external contexts by reordering the text,and integrate external context with original text to build a new NER model.This model can efficiently recognize entities,and reduce the the effect of noisy data on the entities semantic representation.On both Co NLL-2003 and WNUT-16 datasets,the proposed model achieved a satisfactory performance.(2)The method of the scholar academic knowledge network dynamic evolution.The scholar academic knowladge network is incomplete.This paper constructed contextual semantic-guided entity-centric GCN for relation extraction,it can classify the relation between scholar entities.By employing entity alignment method to integrated these relations into the knowledge network,it can implement scholar academic knowledge network dynamic evolution.The relation extraction model consists of the self-enhanced neural network and the full-connected GCN.The selfenhanced neural network can construct sematic-guided context to guide the relation representation between entities.By constructing entity-centric logical adjacency matrix,entities fuse all node semantic features with 1-layer GCN.Experiments show that,the model performed well in the Sem Eval-2010 Task 8 and TACRED public datasets.(3)Scholar academic network service platform.Scholar resources lack unified organization and management on the Internet,and scholar academic networks need to evolve dynamically.The construction of a scholarly web service platform for academics includes an entity identification approach and incremental,evolutionary computation.What‘s more,it can provide scholars with accurate information services such as knowledge integration,academic presentation,and scholar biography generation.48 pictures,13 tables,89 references...