Text Mining For Herb In Prescriptions Of Traditional Chinese Medicine

In recent years,some progress has been made in the research of pharmacology of traditional Chinese medicine(TCM),and a large number of pharmacological literatures of TCM have been accumulated.In this thesis,the pharmacological actions of prescriptions are studied according to the pharmacological literature of TCM,which provides theoretical reference for further understanding the actual pharmacological actions of prescriptions,provides basis for clinical practice,and explores new uses of prescriptions.This thesis takes Chinese core journal literature as the carrier,combines text mining technology to mine the pharmacological actions of single drug,and carries out the pharmacological actions mining of prescriptions based on the pharmacological actions of single drug,takes specific diseases as an example to verify the rationality of prescriptions diseases treatment.Based on above researchs,we build a prototype of literature mining service platform of TCM.Specific works are as follows:1.In order to solve the problem of time-consuming and laborious reading of a large number of pharmacological literatures,this thesis proposes to introduce text mining technology into pharmacological literature of TCM and carry out entity recognition of pharmacological action of TCM.According to the pharmacological types of TCM and the descriptive characteristics of pharmacology in the text,the pharmacological actions are divided into general pharmacological actions and pharmacological mechanisms.Bidirectional LSTM-CRF was used to identify general pharmacological effects,and compared with bidirectional LSTM-Softmax,dictionary-based,rule-based and dictionary-rule combination methods,the recognition accuracy reached 0.9338,0.9292,0.7447,0.6956 and 0.7892,respectively.Using rule-based method to identify pharmacological action mechanism,the accuracy reached 0.8654.2.In order to solve the problem of identifying pharmacological action of prescriptions,this thesis proposes mining pharmacological action of prescriptions based on entity recognition of pharmacological action,extracting general pharmacological action by frequency feature,and constructing pharmacological action mechanism network.Kmeans clustering was used to grouping the pharmacological action mechanism,and Jaccard similarity was used to screen out the core pharmacological action mechanism representing the whole pharmacological action mechanism network.General pharmacological action and core pharmacological action mechanism were selected to represent the pharmacological action of prescriptions.3.In view of the rationality of disease treatment of traditional Chinese medicine prescriptions,this thesis carries out the verification research on the rationality of prescriptions.1)Taking chronic kidney disease as an example,a weighted drug network based on the prescriptions set of medical records was constructed,and disease core drugs were screened by improved mutual information method.The rationality of prescriptions in treating disease was verified by calculating the similarity of the pharmacological action of prescriptions and the pharmacological action of core drugs of the disease;2)The enrichment analysis for TCM ingredients targets and disease-related genes was performed based on the network pharmacology method,further revealed the rationality of prescriptions for diseases.4.Using Eclipse development tools and MySQL database,a prototype of Chinese medicine literature information mining service platform is designed and implemented.The prototype includes data import module,document mining and analysis module,results display and download module.The function of system contains entity recognition of Chinese medicine pharmacological action and mining pharmacological effects of prescriptions.