Research On Construction And Application Of Knowledge Graph For Scientific Publications In Entrepreneurship Domain

Scientific papers are considered one of the most important carriers of knowledge and information in various disciplines and research fields.As they are the most cutting-edge,authoritative,and easiest-to-access knowledge resources in all fields,which embody the outstanding research ideas,theories,and achievements of the scholars.Facing the massive academic resources,how to fully mine the knowledge in scientific papers and provide rich semantic information for the paper retrieval become the important research issues for the development of scientific knowledge management and semantic search.Knowledge graph technology can help solve problems above all.And knowledge graph also helps standardize the management of scientific publications,scholars,academic activities and other scientific entities,which makes researchers find their needed documentation quickly and efficiently.At present,the applied research on the combination of knowledge graph and scientific resource management has been carried out and achieved remarkable results in the industry and academia.Academic search platforms such as Aminer and Microsoft Academic Graph have been supported by the knowledge graph technology.Although they have performed well in academic paper search,academic recommendations,etc.,the mining of relevant entities in scientific papers is not sufficient.Also there is too little metadata of the papers can be searched,and the rich semantic information and domain knowledge in the main body of the papers are not fully understood.This research will dig deeper into the scientific entities and concepts hidden in the papers,and provide researchers with richer knowledge.In this paper,we research on the construction and application of knowledge graph for scientific publications in the entrepreneurship domain.The main contribution of this dissertation can be summarized as follows:1.A construction framework of knowledge graph for scientific publications.First,we define six entities for the knowledge graph,including paper,person,organization,journal,conference and location.Then we define the data schema according to the entities,also including the attributes of entities and relations.2.An introduction to entity and concept recognition model.This model extracts the research object entity,research method,and research theory from papers.This paper mainly adopts the method of combining dictionary and pattern matching algorithm.First we establish geographical entity candidate set and domain knowledge set,then use AC automation for pattern matching.Finally,we get the geographical entities,research methods,research theories extracted from the full text of scientific papers.We also acquire the knowledge hidden inside the papers.3.An automatic semantic annotation algorithm for scientific abstracts.The can be divided into four semantic modules:research background,research purpose,research content and research conclusion.The algorithm extracts sentence features from the three dimensions of category,location and semantics based on the unit of sentence.The abstracts have already been manually annotated.They are trained and tested using models,such as logistic regression and support vector machines.Finally,we obtain pretty good classification results in the case of small-scale corpus.4.A Scientific Entity Knowledge Graph Sharing Platform.The platform is an application of the knowledge graph.The main function are entity search,result display,and full text search based on Elasticsearch.The platform can also visualize knowledge graph of scientific papers.The construction method of knowledge graph for scientific papers proposed in this paper is portable and extensible,and it can be applied to scientific literature management,search and recommendation in any subject area.In the future research,the knowledge graph for scientific papers can be expanded into other scientific entities such as patents,projects,etc.,and will become a scholar hub for scientific resources.