Research On Construction Of Obstetric Knowledge Graph

In order to promote a long-term balanced development of the population,Twochild policy has been overall implemented in 2016,the number of parturients of advanced reproductive age and second-birth parturients is increasing year by year.These two groups are prone to complication of pregnancy,causing a challenge to the obstetrics of medical institutions.The obstetric knowledge Graph can integrate obstetric knowledge well and is understood by the computer,it can be used in optimized treatments and adjunctive therapies,which is the foundation of developing related intelligent applications.However,the construction of medical Knowledge Graph is at the start age.As a consequence of it,construction methods of the obstetric Knowledge Graph are explored in this paper and the obstetric Knowledge Graph is initially constructed.The main research works of this paper are as followings.The construction of schema layer of obstetric Knowledge Graph is completed based on the concept of ontology.Besides,the schema layer is based on Medical Subject Headings and obstetrics professional knowledge and the concept classification system of obstetric is completed under the guidance of domain experts,defining possible relational categories between the knowledge description system of diseases and conceptual entity in obstetric areas and forming a standardized description of obstetric expertise.Schema layer defines the conceptual model of obstetric knowledge graph,which lays the foundation for the construction of data layer.The construction of data layer can be divided into four parts: entity recognition,attribute extraction,attribute fusion and relation extraction.In the entity recognition part,the similarity algorithm of obstetric terms is used to integrate the concepts defined by schema layer and its various expressive forms,creating the dictionary of obstetric terms as the basis for entity recognition of obstetric concepts.In the attribute extraction part,the extraction of disease attributes in obstetrics online resources and maternity textbooks is respectively achieved by creating wrapper and being based on rules and the extraction results are structured in a certain degree.In the attribute fusion part,the Simhash obstetric texts similarity algorithm based on improved TF-IDF is designed and the attribute fusion strategy is proposed to complete the fusion of disease attributes.In the relation extraction part,the relationship pattern is represented by the shortest dependency pattern and the training data are collected by Bootstrapping and artificial methods.Moreover,the extraction of relationships among obstetric conceptual entity is completed by training the model of Support Vector Machine to get the relational classifier.The construction of obstetric Knowledge Graph also needs Protégé.Concepts,entities,attributes and relationships of obstetric Knowledge Graph are established in Protégé,adding constraint conditions to attributes and relationships and importing data by automatically creating Protégé source files.