Research On Generation Of Scientific Research Review Based On Knowledge Graph

Under the background of open science,the explosive growth of scientific papers and other literature resources has greatly exceeded the limit of manual processing,making it increasingly difficult for researchers to accurately obtain relevant information on research topics within a reasonable time.In order to solve the literature acquisition dilemma,scientific research summarization or review oriented to multiple types of data formats has become a hot issue in the academic community in recent years.The existing research review generation research mainly includes automatic text summarization for document data and graph review generation based on literature data.However,limited by the text computing technologies such as natural language processing,the existing scientific research review studies have problems such as limited resource processing scale,unstable quality,insufficient knowledge mining depth,and single display form.There is still a certain distance from artificial effects and it is necessary to seek new technological paths or solutions.With the development of computing technology and the increase in the needs of scientific research users,the generalized summarization is diversifying in terms of core methods and presentation forms.Knowledge graph is a new data infrastructure and knowledge organization form in the era of academic big data.Its semantic standardization and linking ideas can gradually refine the originally unstructured and unrelated rough data into structured and strongly related high-quality knowledge,which can support the multi-angle organization and disclosure of scientific literature structure and subject information shows great application value in data analysis and mining.Taking scientific knowledge graph oriented to thematic data and related scientific literature resources as an important technical support,this dissertation explores several practical scientific issues in the realization of scientific research review based on the knowledge graph.The main research contents and achievements include the following aspects:(1)A scientific research ontology that deeply integrates scientific literature and thematic data is designed to support the fusion organization of multi-source data.Aiming at the problem that the relationship between literature and subject in the existing semantic description model of academic resources is too simple and the thematic data model lacks hierarchical structure,this paper designs a scientific research ontology by combining the logical expression characteristics of terms and concepts in the thematic data,and builds a multi-dimensional semantic description model based on the semantic connectivity path between the literature structure characteristics and the semantic elements of the research topic,which integrated the subject,scientific literature resources and the subject of scientific research activities,providing a semantic basis for the construction of scientific research knowledge graph.(2)A scheme and tool for constructing scientific knowledge graph for scientific literature and thematic data is developed to realize the knowledge extraction and fusion management of multi-source corpus.This paper designs data processing methods covering algorithmic models such as entity disambiguation,literature importance calculation,literature resource subject indexing,abstract move structure recognition,etc.,and develops a tool for knowledge generation,update and management for multi-source heterogeneous corpus data based on the open source ETL framework,thus forming a full-process knowledge graph construction method based on RDF graph data models.The comprehensiveness functions of the tool and one-stop operation characteristics have a certain degree of innovation and strong application value in related technical fields.(3)The storage mechanism of knowledge graph mapping across data models and the calculation of point-edge relationship based on knowledge graph are integrated to realize the efficient storage management and application support of scientific knowledge graph.Based on the mapping relationship between the RDF graph data and the property graph data of knowledge graph,this paper designs an index strategy that supports multiple sub-categories,formulates the storage management method of the scientific knowledge graph and the interface support strategy based on the graph algorithm,and realizes query-oriented semantic parsing and instance matching,and entity and relationship clustering based on sub-graph structure,laying data flow logic for the realization of scientific research summarization.(4)A prototype system of scientific research review based on knowledge graph is constructed to realize query-oriented structured and aggregated knowledge display.This paper designs the architecture and functional modules of a scientific research review system based on knowledge graph,support for a variety of application scenarios such as the clustering of thematic knowledge and literature,the development context of important literature,the evolution analysis of hot topics,and the recommendation of high-impact experts,and combines the knowledge graph and POI to automatically generate the research review document which can be viewed and downloaded,so as to realize the verification of multi-dimensional and structured scientific research review.The comparison with the existing typical review or summarization systems shows that the research review system combined with the graph structure characteristics of the knowledge graph has a certain degree of innovation and breakthrough in data processing scale,operation flexibility and presentation form.