| BackgroundAcne vulgaris is a chronic skin disease that occurs in adolescence and mainly affects the face.It has a high incidence and is detrimental to the physical and mental health of patients.Although there is no such disease name in traditional Chinese medicine,it can belong to the categories of certain diseases in this field.Since ancient times,the understanding of acne vulgaris in traditional Chinese medicine has formed a relatively systematic theoretical system from etiology and pathogenesis to principle-method-recipe-medicines.Based on clinical experience,a expert consensus and considerable experience has been accumulated.Previous studies have summarized and compared the clinical experience on acne vulgaris of individual doctors and schools by means of literature collation,experience summary and data mining.However,due to the dispersion of experience,a large amount of tacit knowledge related to acne vulgaris in implied in TCM clinical experience data is difficult to be excavated and utilized.ObjectiveTo explore the key technologies for the construction of knowledge graph related to TCM diseases,and construct the knowledge graph of TCM for acne vulgaris including theoretical and practical knowledge according to the characteristics of domain knowledge.Literatures of acne vulgaris in the field of TCM serve as the data source.Represent the relationships in TCM about acne vulgaris with the help of the advantages of knowledge graph in association,representation and utilization of knowledge,as well as the reuse of top-level knowledge classification standards and glossaries in TCM and related fields,so as to improve the quality of knowledge service and realize the integration and knowledge discovery of clinical experience from famous doctors in this field.Methods(1)The knowledge levels involved in the knowledge graph were formulated according to the target of knowledge graph of TCM for acne vulgaris.By sorting out the knowledge system of diagnosis and treatment of acne vulgaris in TCM and referring to the existing knowledge classification standards in TCM,the knowledge framework of knowledge graph was formulated.(2)The characteristics and concern of diagnosis and treatment knowledge related to acne vulgaris in TCM were analyzed.Along with the seven-step method of domain ontology construction,the category and term granularity of the ontology were determined,domain terms were collected,both the domain ontology and the domain general,pattern,disease,treatment and Chinese medicinals glossaries of TCM and Western medicine was reused and integrated,the classification system of ontology was constructed,characteristic data properties that meet the need of knowledge representation for acne vulgaris in TCM were defined,and the construction of ontology involved in the knowledge graph was completed.(3)High-quality factual knowledge of acne vulgaris in TCM was selected.Knowledge extraction technologies combined with manual revision and annotation were used,so as to obtain structured data,carry out knowledge fusion between fact knowledge and between fact and ontology knowledge,construct a knowledge graph for acne vulgaris in TCM,select graph database for storage and realize semantic search and other application examples.Results(1)The knowledge graph’s knowledge structure with good interoperability and representation of experience knowledge for famous doctors was constructed.The knowledge graph for acne vulgaris in TCM aimed to the representation and use of famous doctors’ experience,and formulate the knowledge level of the knowledge graph including patterns and facts.The knowledge level supports the representation of theoretical knowledge and clinical experience,which can not only give full play to the semantic advantages of domain ontology,but also promote the complete preservation,analysis and utilization of tacit knowledge in medical records.Analyzing the knowledge system of acne vulgaris in TCM and existing knowledge classification standards,a framework including 10 semantic types,which are disease,pattern,staging and transmission/transmutation,physiological characteristics,symptom and sign,therapeutic principle and method of treatment,drugs,body system,TCM operations approach and physicians,and 9 relationships,which are treatment,analysis,use,inclusion,process,phenomenon representation,characteristics,instantiation,and subClassOf has been formulated.It can support the interoperability requirements of domain cooperation.(2)An ontology that conforms to the characteristics of knowledge representation of acne vulgaris in TCM was constructed.The ontology part of the knowledge graph for acne vulgaris in TCM takes monosemantic terms as the minimum term granularity.Research teases out the terminology of core concepts in TCM and Western medicine,and select the definition which is characterized by TCM knowledge.Domain-specific terms are collected based on it.Class hierarchies of diseases,patterns,therapeutic principle and method of treatment are emphatically constructed.Data properties are defined,including ID,name,synonym,source,source document,correlation,disease diagnosis,expertise,location,frequency,duration,time,sequence,dosage,course of treatment,special decoction and experience.The ontology are edited using protege.(3)A knowledge graph which is able to improve the quality of knowledge service about acne vulgaris in TCM has been realized.Taking the case records of specialists as the data source of factual knowledge,the research has obtained structured data through rule-based knowledge extraction and manual revision and annotation with unified standards.Ontology knowledge was associated with the clinical experience through disambiguation,and knowledge from case records was integrated through alignment.Selecting graph database Neo4j and data import tool Py2neo for storage,a knowledge graph of TCM for acne vulgaris with 5911 entities and 15193 relationships was constructed.It can support semantic search and knowledge discovery in the field of acne vulgaris in TCM,improving the accessibility of domain knowledge.Conclusion(1)The knowledge hierarchy of knowledge graph of TCM for acne vulgaris should take in fact and model knowledge at least,so that the recognized knowledge in the field and the personal experience of famous doctors can be well represented.It is also necessary to set up characteristic types in schema according to the representation needs of disease knowledge.(2)The construction of fine-grained terms should not be emphasized in the domain ontology of acne vulgaris TCM knowledge graph.Constructing an ontology with comprehensive class hierarchies in order to use the combination of the terms in clinical practice will support the flexible combination of terms in the personalized diagnosis and treatment of acne vulgaris better.(3)The disease knowledge graph of TCM can be constructed according to the order of knowledge schema,domain ontology and factual knowledge.During the construction process,attention should be paid to the description of various kinds of knowledge in diseases in the field of TCM,the application objectives of the knowledge graph and the reuse of existing work in the field.The needs and guidance of domain experts should also be much accounted of. |
