Research And Application Of Expert Recommendation Strategy For Paper Review Based On Knowledge Graph

At present,most paper review expert systems use computer-assisted manual methods to complete expert selection tasks,which have shortcomings such as poor recommendation effectiveness and limited knowledge.As a new application in the field of computer science,knowledge graph has been widely studied in the fields of intelligent search,intelligent question answering,and intelligent recommendation since its launch.However,the technology of constructing expert knowledge graph and applying it to expert recommendation in paper review is not yet very mature.Therefore,this article conducts research and analysis on the construction of expert knowledge graph and paper review expert recommendation,constructs an expert knowledge graph,improves a paper review expert recommendation algorithm based on knowledge reasoning,and provides a design scheme for a paper review expert recommendation system.The main research work and achievements include:1.Expert knowledge graph construction work.In terms of entity recognition,a BERTGlo Ve-Bi LSTM-CRF based entity recognition model in the field of computer science was constructed using the large Chinese domain terminology knowledge base of Northeastern University as the entity recognition annotation training set and the abstract of CNKI computer science papers as the training corpus.In terms of relationship extraction,a global similarity algorithm is proposed by integrating methods such as word level,semantic distance measurement,and semantic calculation.Experimental analysis shows that compared to traditional models,this entity recognition model has a certain improvement effect in the accuracy,recall,and F1 value of recognition tasks.2.An improved expert recommendation algorithm for paper review based on knowledge reasoning is proposed.Based on the representation results of the paper,the constructed expert knowledge graph is used to achieve entity chain indexing and knowledge reasoning,and to construct a domain candidate set.A domain authority evaluation algorithm based on time decay factor and trust relationship is proposed.The main principle is to introduce a time decay factor to refine the impact of evaluating experts’ authority on published documents at different times;By introducing trust relationships,we measure the consistency of review experts’ review work in the relevant field,and dynamically adjust it through review feedback.Experimental analysis shows that the proposed evaluation algorithm has effectively improved the recommendation accuracy for expert selection tasks in papers.3.Based on the research results of this article,using Vue framework,Spring Boot framework,and My Batis framework,a design scheme of a paper review expert recommendation system is proposed,and its prototype system is implemented.The system includes four core functions: graph construction,expert recommendation,expert management,and application management.The test results show that the system meets the expected requirements and functions.