Research On Construction Method Of Domain Knowledge Base Based On Probability Graph Model

Knowledge base on traditional Chinese medicine(TCM)domain is an important way to store and express knowledge of TCM,and provides technical support for clinical research and decision-making of modern medicine,therefore,the construction of knowledge base of TCM has become a research hotspot at present.Treatment based on syndrome differentiation is the basic principle for doctors to recognize and treat diseases,although the basic theories followed by medical doctors are the same or similar,there are some differences in the specific diagnosis and treatment.Therefore,understanding and grasping the differences of TCM syndrome differentiation and treatment is an important basis for the inheritance and development of TCM personalized knowledge.At present,most of the TCM knowledge bases that have been constructed so far only have the integration of the clinical practice knowledge of famous and old TCM doctors,which cannot reflect the difference of syndrome differentiation and treatment,and lack of the discovery of doctors’ individual diagnosis and treatment experience and rules.Based on the theory of probability graph model,a knowledge base that can reflect the difference of syndrome differentiation and medication was constructed by adding the attributes of "doctor" and "value" to the relationship.It mainly includes the following aspects:1)The "relationship strength" is expressed quantitatively to reflect the difference of syndrome differentiation and treatment.Firstly,according to the existing infertility data,the Page Rank algorithm is used to calculate the strength of the relationship between entities,which is easy to be counted.Then,for the relationship strength which is difficult to be calculated by statistical methods,probabilistic soft logic is used to perform interpretable reasoning based on the results of Page Rank algorithm.Finally,comparative experiments are conducted on infertility data sets to prove the validity of probabilistic soft logic reasoning relationship strength.2)A knowledge fusion model based on relational graph attention network was proposed to fuse the personalized knowledge base of infertility among multiple doctors.Firstly,in the graph attention network,the influence of the relationship on the entity information is not considered when the neighbor entity information is propagated and aggregated,the relational graph attention network is designed,and the information of edge and node in entity node’s neighborhood is weighted and aggregated,so as to obtain the rich semantic representation of entity node and realize the fusion of multiple knowledge bases.Then,compared with other knowledge fusion models,the validity of the model is verified.3)In order to intuitively reflect the difference of syndrome differentiation and treatment,knowledge about infertility was stored in Neo4 j,forming a infertility knowledge graph with1131 entity nodes and 28903 relationships,and visualizing the difference of syndrome differentiation results and drug use respectively.Figure 21;Table 26;Reference 64.