A method and device for constructing a knowledge graph based on Treatise on Febrile Diseases

Abstract

The present application relates to a method and device for constructing a knowledge graph based on Treatise on Febrile Diseases. The method comprises: analyzing and mining the original text of Treatise on Febrile Diseases, constructing a conceptual system of disease, syndrome, symptom, method and prescription, and constructing a conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between the five determined by the system; extracting knowledge from the original text of Treatise on Febrile Diseases based on the framework, and constructing a conceptual hierarchy structure tree according to the hierarchical relationship; and extracting entities, entity attributes and entity relationships from the original text of Treatise on Febrile Diseases based on the conceptual hierarchy structure tree; and storing the knowledge triple set obtained after knowledge fusion in a Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases. The method can realize the digitization of the knowledge system of Treatise on Febrile Diseases, and can clearly and efficiently obtain the connection between "disease-syndrome-symptom-method-prescription", which helps learners of Treatise on Febrile Diseases to better and faster master the knowledge system of Treatise on Febrile Diseases.

Description

Technical Field

[0001] The present application relates to the fields of traditional Chinese medicine and computer information technology, and in particular to a method and device for constructing a knowledge graph based on Treatise on Febrile Diseases. Background Art

[0002] Knowledge graph technology is one of the key technologies supporting the development of artificial intelligence. It describes concepts, entities and their relationships in the objective world in a structured form, expresses Internet information in a form that is closer to the human cognitive world, and provides a better ability to organize, manage and understand massive Internet information.

[0003] In recent years, with the proposal and practice of smart medicine and precision medicine, as well as the development of knowledge graph technology, it has become possible to build cheap, efficient and accurate medical diagnosis systems and decision support systems, and medical knowledge graphs have shown increasingly important application value. The application of knowledge graphs in the field of traditional Chinese medicine is currently mainly focused on the representation and integration of traditional Chinese medicine knowledge, as well as shallow knowledge reasoning.

[0004] Ruan Tong and others cooperated with Shanghai Shuguang Hospital to build a knowledge graph of traditional Chinese medicine, including disease database, syndrome database, symptom database, Chinese herbal medicine database and prescription database, and conducted TCM Q&A and assisted prescription based on the knowledge graph. At present, the system has a lot of manual intervention in the automatic extraction and annotation of data. In terms of knowledge fusion and quality assessment, it also relies more on the intervention of human experts, and the knowledge reasoning ability needs to be improved. Yu Tong and others collected and sorted out knowledge resources related to TCM health preservation based on the ontology of TCM health preservation and constructed a knowledge graph of TCM health preservation. In addition, the Institute of Information of Traditional Chinese Medicine of the China Academy of Chinese Medical Sciences opened TCM medical case knowledge graph, TCM health preservation knowledge graph, TCM classic prescription knowledge graph, TCM beauty knowledge graph and other TCM knowledge graphs to the public through the knowledge service platform. These knowledge graphs are all realized by expanding the existing domain ontology. From the effect of page interaction, they can provide users with knowledge search services in a visual way, but they cannot realize more complex user queries and knowledge reasoning. Summary of the invention

[0005] Based on this, it is necessary to provide a method and device for constructing a knowledge graph based on Treatise on Febrile Diseases to address the above technical issues.

[0006] A method for constructing a knowledge graph based on Treatise on Febrile Diseases, the method comprising:

[0007] The original text of Treatise on Febrile Diseases was analyzed and mined, and a conceptual system of disease, syndrome, symptom, method and prescription was established from the three aspects of theory, method and prescription. The relationship between disease, syndrome, symptom, method and prescription was determined based on the conceptual system, and a conceptual relationship description framework of Treatise on Febrile Diseases was constructed based on the relationship between disease, syndrome, symptom, method and prescription.

[0008] According to the concept relationship description framework of Treatise on Febrile Diseases, with reference to predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a concept hierarchy structure tree is constructed according to the hierarchical relationship; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships.

[0009] According to the concept nodes, the semantic relations and the constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the original text of Treatise on Febrile Diseases to obtain a set of knowledge triples.

[0010] The knowledge triple set is subjected to knowledge fusion to obtain a knowledge triple set after knowledge fusion and a synonym library of Treatise on Febrile Diseases.

[0011] The knowledge triple set is stored in the Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases.

[0012] A knowledge graph construction device based on Treatise on Febrile Diseases, the device comprising:

[0013] The module for constructing the conceptual relationship description framework of Treatise on Febrile Diseases is used to analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, determine the relationship between disease, syndrome, symptom, method and prescription based on the conceptual system, and construct the conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription.

[0014] The pattern layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to extract knowledge from the original text of Treatise on Febrile Diseases based on the concept relationship description framework of Treatise on Febrile Diseases and with reference to predetermined technical specifications, and to construct a concept hierarchy structure tree according to the hierarchical relationship; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships.

[0015] The data layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to perform entity extraction, entity attribute extraction and entity relationship extraction on the original text of Treatise on Febrile Diseases according to the concept nodes, the semantic relationships and the constraint rules to obtain a set of knowledge triples; perform knowledge fusion on the set of knowledge triples to obtain a set of knowledge triples after knowledge fusion and a synonym library of Treatise on Febrile Diseases; store the set of knowledge triples using the Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases.

[0016] The above-mentioned method and device for constructing a knowledge graph based on Treatise on Febrile Diseases analyzes and mines the original text of Treatise on Febrile Diseases, constructs a conceptual system of disease, syndrome, symptom, method and prescription, and constructs a conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription determined by the conceptual system; according to the conceptual relationship description framework of Treatise on Febrile Diseases, with reference to predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a conceptual hierarchy structure tree is constructed according to the upper and lower relationships; and according to the conceptual hierarchy structure tree, entity extraction, entity attribute extraction, entity relationship extraction and knowledge fusion are performed on the original text of Treatise on Febrile Diseases, and the knowledge triple set obtained after knowledge fusion is stored in a Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases. The application of this method can realize the digitization of the knowledge system of Treatise on Febrile Diseases, can clearly and efficiently obtain the relationship between "disease-syndrome-symptom-method-prescription", and help learners of Treatise on Febrile Diseases to better and faster master the knowledge system of Treatise on Febrile Diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a process of constructing a knowledge graph based on Treatise on Febrile Diseases in one embodiment;

[0018] Figure 2 The classification of diseases in Treatise on Febrile Diseases in one embodiment;

[0019] Figure 3 is a conceptual hierarchy of six meridian diseases and syndromes in an embodiment;

[0020] Figure 4 is a conceptual hierarchy of six-channel pathological syndromes in another embodiment;

[0021] Figure 5 The classification structure of the treatment principles and methods of "Treatise on Febrile Diseases" in another embodiment;

[0022] Figure 6 This is an example of a knowledge graph (part) of "Treatise on Febrile Diseases" in another embodiment;

[0023] Figure 7 A schematic diagram of a process of constructing a knowledge graph based on Treatise on Febrile Diseases in another embodiment;

[0024] Figure 8 is a flow chart of step 104 in another embodiment;

[0025] Fig. 9 is a flow chart of step 200 in another embodiment;

[0026] Fig.10 A knowledge graph of Taiyang disease syndrome in Treatise on Febrile Diseases in another embodiment;

[0027] Fig.11 This is a knowledge graph of the symptoms of "aversion to cold" in Treatise on Febrile Diseases in another embodiment;

[0028] Fig.12 This is a knowledge graph of medical records of mild depression syndrome in Treatise on Febrile Diseases in another embodiment;

[0029] Fig.13 This is a structural block diagram of a device for constructing a knowledge graph based on Treatise on Febrile Diseases in another embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0031] The book discusses typhoid fever, stroke, febrile disease, wind-heat and other contents, but from the main content, it focuses on the diseases caused by the human body being affected by wind-cold evil, as well as the differentiation rules of combined diseases and concurrent diseases. The book has 10 volumes and 22 chapters, which can be divided into three parts, namely the first 4 chapters, the middle 10 chapters, and the last 8 chapters. The first 4 chapters mainly talk about the characteristics and identification methods of normal pulse and pathological pulse, and summarize the influence of exogenous febrile diseases and seasons on the human body, as well as spasmodic dampness and sunstroke caused by exogenous evil; the middle 10 chapters mainly discuss the six meridian diseases and their variants, which are the focus of study and research of doctors in all dynasties, and are also the source of the content of the "Treatise on Febrile Diseases" course in contemporary colleges of traditional Chinese medicine; the last 8 chapters classify and explain the treatment methods that can and cannot be used, and most of their contents are scattered in the middle 10 chapters.

[0032] In one embodiment, Figure 1 As shown, a method for constructing a knowledge graph based on Treatise on Febrile Diseases is provided, and the method should include the following steps:

[0033] Step 100: Analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, determine the relationship between disease, syndrome, symptom, method and prescription based on the conceptual system, and construct a conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription.

[0034] Specifically, the original text of Treatise on Febrile Diseases involved in this method is the middle 10 chapters of Treatise on Febrile Diseases (a total of 398 articles). Since many articles in the middle 10 chapters involve pulse conditions, seasons and solar terms, and the permissibility of sweating, vomiting, and purgation, the contents of the other 12 chapters will also be referenced in the process of knowledge representation and formalization of these articles.

[0035] The data sources of this method are mainly "Annotated Treatise on Febrile Diseases" (edited by Liu Duzhou and published by People's Medical Publishing House), which is based on the reprint of Zhao Kaimei in the Ming Dynasty, "Selected Readings from Treatise on Febrile Diseases (New Century 4th Edition)", a national higher education planning textbook for the traditional Chinese medicine industry, and "Terms of Clinical Diagnosis and Treatment of Traditional Chinese Medicine (Revised Edition)" (divided into three parts: diseases, syndromes, and treatment methods) issued by the State Administration of Traditional Chinese Medicine in 2009.

[0036] The concept classification of diseases in "Treatise on Febrile Diseases": including exogenous diseases and miscellaneous diseases, among which exogenous diseases can be divided into two categories: typhoid fever and febrile diseases (including wind-heat). "Treatise on Febrile Diseases" mainly discusses the diagnosis and treatment of typhoid fever, that is, exogenous febrile diseases caused by the human body's exposure to wind-cold evil spirits, including six meridian diseases, cholera, yin-yang changes, and fatigue after recovery. Figure 2 shown.

[0037] The six meridians are the abbreviations of Taiyang, Yangming, Shaoyang, Taiyin, Shaoyin and Jueyin by later generations of doctors. In the six meridians differentiation, the three yin and three yang are pathological concepts, including the location of the disease, the nature of the disease, and the course of the disease. It is actually a summary of the disease. Table 1 is the meaning of the three yin and three yang of the six meridian diseases summarized by Professor Hao Wanshan. Therefore, in the process of constructing the digitalization of the knowledge system of "Treatise on Febrile Diseases", the present invention will include the knowledge of the location of the disease, the nature of the disease, the course of the disease, etc. for each meridian disease.

[0038] Table 1 Key points for syndrome differentiation of six meridian diseases

[0039]

[0040] Conceptual classification of syndromes in Treatise on Febrile Diseases: From the perspective of facilitating digital processing, the six meridian diseases and concurrent syndromes and the six meridian pathological syndromes are classified into different conceptual classification systems, and the conceptual classification methods of the planned textbook Selected Readings from Treatise on Febrile Diseases (New Century 4th Edition) are selected and discarded, so that the six meridian diseases and concurrent syndromes are classified into the systematic syndrome differentiation category, and the variant syndromes of the six meridian diseases, cholera syndrome, yin-yang change syndrome, fatigue syndrome after recovery, and similar syndromes are classified into the non-systematic syndrome category. In addition, since modern clinical practice is based on syndrome differentiation, it does not have much clinical significance to judge the combined disease, concurrent disease or dual infection in the six meridian diseases, so these three situations are collectively referred to as "six meridian disease concurrent syndromes", and the specific classification is as follows: Figure 3 and Figure 4 shown.

[0041] Conceptual classification of symptoms in "Treatise on Febrile Diseases": In order to eliminate ambiguity in the understanding of TCM symptom terminology and facilitate the digital processing of TCM symptoms and the construction of symptom-based digital syndrome differentiation methods, this article refers to the knowledge framework of the diagnosis method part of the planned textbook "Diagnosis of Traditional Chinese Medicine (New Century 4th Edition)" and the classification method of clinical syndromes in "Syndrome Differentiation of Syndromes and Elements" written by Mr. Zhu Wenfeng (this book defines "syndrome" as the external symptoms of syndrome, referring to all evidence such as symptoms and signs that are intrinsically related to a specific syndrome). From the four aspects of inspection, auscultation, questioning and palpation, the symptoms involved in the original text of "Treatise on Febrile Diseases" are divided into 4 major categories and 21 minor categories, as shown in Table 2.

[0042] Table 2 Classification of symptoms in Treatise on Febrile Diseases

[0043]

[0044]

[0045] In the "Symptom Examples" column of Table 2, the terms in brackets after the symptoms are synonyms. Although the symptoms in the article of "Treatise on Febrile Diseases" do not involve tongue images, considering that the digital syndrome differentiation model constructed in this article is oriented to modern clinical applications, tongue images are an important basis for understanding the condition and syndrome differentiation and treatment, so they are still retained.

[0046] Classification of the concepts of treatment principles and methods in "Treatise on Febrile Diseases": On the basis of "Huangdi Neijing", "Treatise on Febrile Diseases" inherits and develops a variety of treatment principles of traditional Chinese medicine. In terms of treatment methods, basic treatment methods, commonly used therapies and combination therapies have been developed. Basic treatment methods include sweating, vomiting, purgation, harmonization, warming, clearing, supplementing and eliminating methods; commonly used therapies include drug therapy, moxibustion therapy, acupuncture therapy, diet therapy, etc.; combination therapies include drug and acupuncture combination, drug and moxibustion combination, acupuncture and moxibustion combination, drug and food combination and other combination therapies. For this reason, based on Figure 5 The classification structure of treatment principles and methods shown assigns a treatment principle category, treatment method category and therapy category to the articles related to treatment in "Treatise on Febrile Diseases".

[0047] Conceptual classification of prescriptions in Treatise on Febrile Diseases: According to the classification method of New Compilation of Treatise on Febrile Diseases by Mr. Liu Duzhou, all prescriptions are divided into 18 categories, and two attributes, "indications" and "related articles", are set for each prescription to enhance the effect of "understanding the symptoms from the prescriptions and understanding the differentiation from the symptoms". Table 3 shows the specific classification method of 112 prescriptions.

[0048] Table 3 Classification of classical prescriptions

[0049]

[0050]

[0051] The relationships among disease, syndrome, symptom, method and prescription include: disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship, syndrome-method relationship, syndrome-prescription relationship, and method-prescription relationship.

[0052] The conceptual relationship description framework of "Treatise on Febrile Diseases" is constructed based on seven major types of relationships: "disease-disease" relationship, "disease-syndrome" relationship, "disease-symptom" relationship, "syndrome-symptom" relationship, "syndrome-method" relationship, "syndrome-prescription" relationship, and "method-prescription" relationship.

[0053] Step 102: Based on the concept relationship description framework of Treatise on Febrile Diseases and referring to the predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a concept hierarchy structure tree is constructed according to the hierarchical relationship; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships.

[0054] Concept hierarchy structure tree: 25 representative concept nodes, 27 semantic relationships, and 44 main constraint rules between concept nodes and semantic relationships.

[0055] The concept hierarchy tree is a hierarchical structure composed of attribute values ​​and concepts in the database according to different levels of abstraction. It is a generalized semantic description of the attribute. The high-level concepts in the tree are generalizations of the bottom-level concepts. The root of the tree is the most general description of the possible attribute values ​​of the attribute, and the leaves are the possible attribute values ​​of the attribute.

[0056] Step 104: According to the concept nodes, semantic relations and constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the original text of Treatise on Febrile Diseases to obtain a set of knowledge triples.

[0057] Specifically, according to concept nodes, semantic relationships and constraint rules, the method of "manual annotation + machine extraction + expert review" is adopted to perform entity extraction, entity attribute extraction and entity relationship extraction on the original text of Treatise on Febrile Diseases to obtain a set of knowledge triples.

[0058] Step 106: Perform knowledge fusion on the knowledge triple set to obtain the knowledge triple set after knowledge fusion and the synonym database of Treatise on Febrile Diseases.

[0059] Knowledge fusion of knowledge triple sets is done manually to complete entity alignment.

[0060] Step 108: Use the Neo4j graph database to store the knowledge triple set and complete the construction of the knowledge graph of Treatise on Febrile Diseases.

[0061] Specifically, since the knowledge graph consists of a data layer and a model layer, the concept hierarchy structure tree here is the model layer of the knowledge graph, and the set of knowledge triples after knowledge fusion is the data of the knowledge graph data layer.

[0062] The knowledge graph can be formally represented as a pattern graph G s 、Data graph G d , the relationship between the two, that is: G = <G s ,G d ,R>. Pattern diagram G s = <N s ,E s >, where: N s The node set representing the concepts in the knowledge graph, E s A set of edges representing the semantic relationship between concepts in a knowledge graph. d = <N d ,E d >, where: N d A set of nodes representing instances (i.e. entities) and strings (used to represent attribute values ​​of instances) in the knowledge graph, E d A set of edges representing the semantic relationship between instances in the knowledge graph. d and pattern diagram G s The relationship R between them is composed of rdf:type to indicate the subordinate relationship between the instance and the concept to which it belongs. Figure 6 This is an example of the knowledge graph (part) of "Treatise on Febrile Diseases".

[0063] In the above-mentioned knowledge graph construction method based on Treatise on Febrile Diseases, the method analyzes and mines the original text of Treatise on Febrile Diseases, constructs a conceptual system of disease, syndrome, symptom, method and prescription, and builds a conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription determined by the conceptual system; according to the conceptual relationship description framework of Treatise on Febrile Diseases, with reference to predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a conceptual hierarchy structure tree is constructed according to the upper and lower relationships; and according to the conceptual hierarchy structure tree, entity extraction, entity attribute extraction, entity relationship extraction and knowledge fusion are performed on the original text of Treatise on Febrile Diseases, and the knowledge triple set obtained after knowledge fusion is stored in a Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases. The application of this method can realize the digitization of the knowledge system of Treatise on Febrile Diseases, and can clearly and efficiently obtain the relationship between "disease-syndrome-symptom-method-prescription", which helps learners of Treatise on Febrile Diseases to better and faster master the knowledge system of Treatise on Febrile Diseases.

[0064] In one embodiment, if Figure 7 As shown, the method also includes:

[0065] Step 110: According to the concept nodes, semantic relations and constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the medical records of famous typhoid fever doctors to obtain knowledge triples of the medical records of famous typhoid fever doctors.

[0066] The method of extracting knowledge from famous medical records of typhoid fever is the same as the method of extracting knowledge from the original text of Treatise on Febrile Diseases.

[0067] Step 112: Map the entity names in the knowledge triples of famous typhoid fever medical records to standard entities according to the synonym database, and use a statistical method to calculate the entity similarity; the entity similarity is a weighted sum of character similarity and semantic similarity.

[0068] For the knowledge from famous medical records of typhoid fever, we use a mixed method of rules and statistics to achieve entity alignment, and submit similar entities to TCM experts for review to ensure the quality of knowledge fusion. First, the entity names in the triples are mapped to standard entities with the help of a synonym library, and then the entity similarity is calculated based on a statistical method to achieve entity alignment, including character similarity calculation and semantic similarity calculation. The entity similarity calculation method is shown in formula (1):

[0069] sim(A,B)=(1-a)*J(A,B)+a*sim s (A,B) (1)

[0070] Among them, sim(A,B) is the similarity between entity A and entity B, J(A,B) is the character similarity, sim s (A, B) is the semantic similarity, a is the weight, and preferably, the weight a=0.8.

[0071] Step 114: Submit the entity with the highest entity similarity greater than the preset threshold as a candidate alignment entity to the TCM experts for review; if the review is passed, merge the attributes and relationships of the two entities, and establish a synonym relationship between the entity and the standard entity in the disease-syndrome-symptom-method-prescription knowledge graph of Shanghan Lun; if the review is not passed, add the entity as a new entity to the disease-syndrome-symptom-method-prescription knowledge graph of Shanghan Lun. Preferably, the preset threshold is 0.8.

[0072] In one embodiment, the character similarity in step 112 is calculated using the Jaccard distance between two entity character strings, and the expression of the character similarity is:

[0073]

[0074] Among them, || is the number of characters to be calculated, |A∩B| is the character intersection of entity A and entity B, and |A∪B| is the character union of entity A and entity B.

[0075] Semantic similarity is calculated using the cosine similarity between two entities. The expression of semantic similarity is:

[0076]

[0077] Among them, sim s (A,B) is the semantic similarity between entity A and entity B, (x1,x2,…,x n ) is the vector representation of entity A, (y1,y2,…,y n ) is the vector representation of entity B.

[0078] In one embodiment, step 100 includes: analyzing and mining the original text of Treatise on Febrile Diseases, establishing a conceptual classification system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, and collecting relevant vocabulary from the five aspects of disease, syndrome, symptom, method and prescription as standard terms and adding them into the term dictionary; according to the conceptual classification system of disease, syndrome, symptom, method and prescription, defining the relationship between disease, syndrome, symptom, method and prescription by using conceptual relationship and attribute relationship, the relationship between disease, syndrome, symptom, method and prescription includes: disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship, syndrome-method relationship, syndrome-prescription relationship and method-prescription relationship; the conceptual relationship is represented by a triple <concept A, relationship R, concept B>, wherein the types of relationship R include: one-to-one, one-to-many and many-to-many; the attribute relationship is represented by a triple <concept A, attribute B, attribute value C> , where the attribute value type is a number or a string, and the attribute values ​​include: name, duration, severity, location, and nature of the disease; based on the disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship, syndrome-method relationship, syndrome-prescription relationship, and method-prescription relationship, a conceptual relationship description framework of Treatise on Febrile Diseases is constructed.

[0079] Specifically, this embodiment defines a total of 27 concept relationships and 36 attribute relationships. The term dictionary also includes the content of the 2006 edition of the National Standard for Basic Theoretical Terminology of Traditional Chinese Medicine.

[0080] "Disease-disease" relationship: Generally, it is divided into "transmission" relationship and "concurrence" relationship. In the six meridian diseases, the evil qi enters the meridian from this meridian, which is called transmission. The "disease-disease" relationship of the three situations of combined disease, concurrent disease, and double infection is called "concurrence" relationship.

[0081] The "disease-syndrome" relationship is a one-to-many relationship, and the relationship between disease and syndrome is a conceptual inclusion relationship. For example, Taiyang disease is divided into Taiyang disease meridian syndrome and Taiyang disease visceral syndrome, and Taiyang disease meridian syndrome can be divided into stroke superficial deficiency syndrome, typhoid fever superficial excess syndrome, superficial depression mild syndrome, etc.

[0082] The "disease-symptom" relationship is a many-to-many relationship. Because a symptom may appear in the outlines of multiple meridian diseases, a meridian disease contains multiple symptoms.

[0083] The "syndrome-symptom" relationship is subdivided into the "main symptom" relationship and the "concomitant symptom" relationship, both of which are many-to-many relationships.

[0084] The "symptom-treatment" relationship is a many-to-many relationship. For a specific disease or syndrome, multiple treatment methods may be required. On the other hand, one treatment method can be applied to multiple diseases or syndromes.

[0085] The "symptom-prescription" relationship is a many-to-one relationship. For a specific disease syndrome, the corresponding prescription is fixed. On the other hand, a prescription has multiple indications and can be used for multiple diseases syndromes.

[0086] The "method-prescription" relationship is a one-to-many relationship. For a specific prescription, the corresponding treatment method is fixed. On the other hand, a treatment method can correspond to multiple prescriptions.

[0087] In addition to the relationship between concepts, it is also necessary to consider the relationship between the disease and time (duration of the attack, solar terms, seasons), and quantitative attributes such as the severity, strength, and urgency of the symptoms. Table 4 shows the relationship description framework of syndrome concepts using the Taiyang Stroke Syndrome as an example.

[0088] Table 4 The framework for describing the relationship between syndrome concepts in Treatise on Febrile Diseases (taking Taiyang Stroke Syndrome as an example)

[0089]

[0090] It should be pointed out that the Taiyang stroke syndrome in the above case represents a disease syndrome concept, which contains multiple entities (i.e. specific disease syndromes). In order to distinguish these disease syndromes and give a definite syndrome differentiation result for each medical case, this article follows the provisions of the textbook "Selected Readings of Treatise on Febrile Diseases (New Century 4th Edition)" and names these specific disease syndromes with decoction syndromes, that is, using the method of "disease syndrome concept + prescription syndrome" (e.g. Taiyang stroke syndrome + Guizhi decoction syndrome) to give the syndrome differentiation result.

[0091] In one embodiment, if Figure 8 As shown, step 104 specifically includes:

[0092] Step 200: Under the guidance of TCM experts, based on concept nodes, semantic relationships and constraint rules, and with reference to pre-selected existing compilations of Treatise on Febrile Diseases, the original text of Treatise on Febrile Diseases is converted into a knowledge form to obtain multiple knowledge tables; each cell content in the knowledge table has a clear scope, and synonyms and symptom types are also marked with special characters.

[0093] Step 202: In each cell of the knowledge table, with the help of special characters, the NLPIR-ICTCLAS Chinese word segmentation system is used to implement Chinese word segmentation and complete entity extraction in five aspects: disease, syndrome, symptom, method, and prescription.

[0094] Step 204: According to the semantic relationship and constraint rules, a specific entity type is assigned to each cell of the knowledge table, as well as a specific relationship type between cells, to obtain the knowledge table rules.

[0095] Step 206: Based on the term dictionary and knowledge table rules, a set of knowledge triples is extracted from all knowledge tables to complete the extraction of entity relationships in the five aspects of disease, syndrome, symptom, method, and prescription.

[0096] Specifically, since the content of each cell in these "Treatise on Febrile Diseases" knowledge tables has a clear range, and synonyms and symptom types are also marked with special characters, we adopt a dictionary- and rule-based entity and relationship simultaneous extraction method.

[0097] Each cell in the knowledge table of "Treatise on Febrile Diseases" is assigned a specific entity type, as well as a specific relationship type between cells. The specific meanings are shown in Tables 5 and 6.

[0098] Table 5 Entity and attribute representation methods of the knowledge table of Treatise on Febrile Diseases

[0099]

[0100] In Table 5, the italic characters are fixed contents, which refer to specific meanings. The cell contents referred to by cell01 to cell07 consist of knowledge in five aspects: disease, syndrome, symptom, method and prescription.

[0101] Table 6 List of implicit relationships in knowledge cells of Treatise on Febrile Diseases

[0102]

[0103]

[0104] Based on the constructed terminology dictionary of "Treatise on Febrile Diseases" and the above-mentioned knowledge table rules, 3142 triples were extracted from 318 original knowledge tables of "Treatise on Febrile Diseases", and the relationship extraction of five aspects of "Treatise on Febrile Diseases" was completed.

[0105] In one embodiment, if Fig. 9 As shown, step 200 specifically includes:

[0106] Step 300: For all the texts in Treatise on Febrile Diseases involving disease differentiation and syndrome differentiation, when each text involves only one diagnostic conclusion, knowledge is extracted from the original disease syndrome, treatment history, current disease syndrome, etiology and pathogenesis, syndrome differentiation results, treatment plan and plan coding to obtain a knowledge table corresponding to the text in Treatise on Febrile Diseases; each term in the knowledge type of the current disease syndrome and treatment plan in the knowledge table indicates the concept node to which it belongs.

[0107] For example, Article 12 of the original text of Treatise on Febrile Diseases states, "When a person is struck by a wind in the sun, the yang is floating and the yin is weak. When the yang is floating, heat will spontaneously occur; when the yin is weak, sweat will spontaneously come out. If the person is shivering with aversion to cold, aversion to wind, fever, nasal congestion and dry retching, Guizhi Decoction is the main treatment." This original text has only one diagnosis result, so it does not need to be split. This article will extract knowledge from the original text 12 articles in the manner of Table 7.

[0108] Table 7 Example 1 of knowledge extraction from the original text of Treatise on Febrile Diseases

[0109]

[0110] Each term in the knowledge types such as current disease syndrome and treatment plan in Table 7 indicates the semantic type to which it belongs. The naming conventions of entities such as diseases, syndromes, symptoms, treatment methods, therapies, prescriptions, etc. involved in the knowledge extraction process are processed as follows:

[0111] (a) Disease name and syndrome name: in accordance with the national standards "Classification and Codes of TCM Diseases and Syndromes" (2020 revised edition) and "Terms of Clinical Diagnosis and Treatment of TCM" (2020 revised edition);

[0112] (b) Symptom name: Based on the "Terminology Standards for Common Clinical Symptoms in Traditional Chinese Medicine (Revised)" written by Professor Li Jingbo;

[0113] (c) Names of treatment methods and therapies: Based on the national higher education planning textbook for the traditional Chinese medicine industry, Selected Readings from Treatise on Febrile Diseases (New Century 4th Edition);

[0114] (d) Name of prescription: Based on the Song Dynasty edition of Treatise on Febrile Diseases reprinted by Zhao Kaimei in the Ming Dynasty.

[0115] In the event that the above national standards and works do not define relevant terms, the national higher education planning textbooks for the traditional Chinese medicine industry, "Selected Readings from Treatise on Febrile Diseases (New Century 4th Edition)" and "Annotated Treatise on Febrile Diseases" (1991, People's Medical Publishing House) shall prevail.

[0116] In addition, if there are multiple synonyms under the above regulations, they will be stored in the synonym table. For example, the original text of Article 12 of the original book involves "yang floating and yin weak" and "stroke superficial deficiency syndrome", which are all stored in the synonym table.

[0117] Regarding the regulations on "scheme coding", since words such as "maintains the treatment", "should", "then give", "can be given", "cannot", and "must not" appear when using a certain prescription in the original text of Treatise on Cold Damage Disorders, which represent four different levels of meanings (the first three meanings represent that it can be used, with the degree decreasing layer by layer, and the fourth meaning indicates that it is prohibited from use), therefore, "scheme coding" is adopted in the knowledge type of Table 8 to represent these four different levels of meanings. For this reason, when the words "××× decoction is the main treatment" appear in the original text, the field value is 1; if it is "should", the field value is 2; if it is "then give", "can be given", etc., the field value is 3; if it is "cannot" or "must not", the field value is 4.

[0118] Step 302: When a single scripture involves multiple diagnostic conclusions, split this scripture into multiple scriptures so that each of the split scriptures contains only one diagnostic conclusion, and perform extraction according to the knowledge extraction method for scriptures that only involve one diagnostic conclusion to obtain the knowledge table corresponding to the Treatise on Cold Damage Disorders scriptures.

[0119] For example, in the 15th article of the original text of Treatise on Cold Damage Disorders, "For a patient with taiyang disease, if purging is applied and there is qi surging upward, Cinnamon Twig Decoction can be given, using the previous method. If there is no qi surging upward, it must not be given." Since there are different treatment methods for two disease conditions in this original text, first split this original text into two scriptures numbered "15-1" and "15-2", and then perform knowledge extraction according to the methods in Table 8 and Table 9.

[0120] The content of the scripture numbered "15-1" is: "For a patient with taiyang disease, if purging is applied and there is qi surging upward, Cinnamon Twig Decoction can be given." The method for extracting knowledge from it is shown in Table 8.

[0121] Table 8 Example Two of Knowledge Extraction from the Original Text of Treatise on Cold Damage Disorders

[0122] Serial number Type of knowledge Knowledge Content 1 Original disease Sun disease 2 Treatment history The method of treatment (treatment method + mistreatment) 3 Current disease Consciousness of upward flow of qi (main symptom) 4 Etiology and pathogenesis Misuse of the method of purgation will not eliminate the external evil, but will only hurt the internal Qi, but the positive Qi will be sufficient. 5 Syndrome differentiation results Sunstroke syndrome + Guizhi Decoction syndrome 6 Treatment options Sweating therapy (treatment method), Guizhi Decoction (prescription) 7 Scheme Code 3

[0123] The content of the scripture numbered "15-2" is: "For a patient with taiyang disease, if purging is applied and there is no qi surging upward, Cinnamon Twig Decoction must not be given." The method for extracting knowledge from it is shown in Table 9.

[0124] Table 9 Example Three of Knowledge Extraction from the Original Text of Treatise on Cold Damage Disorders

[0125]

[0126]

[0127] Step 304: Naming specifications for disease entities, syndrome entities, symptom entities, treatment method entities, therapy entities, and prescription entities involved in the knowledge extraction process according to corresponding preset criteria.

[0128] Specifically, for the standardization of disease names and syndrome names, refer to the national standards "Classification and Codes of Chinese Medicine Diseases and Syndromes" (2020 revised edition), "Terms of Clinical Diagnosis and Treatment of Chinese Medicine" (2020 revised edition), "Selected Readings of Treatise on Febrile Diseases (New Century 4th Edition)", and "Annotated Treatise on Febrile Diseases". 15 disease names and 299 syndrome names (including syndrome names in the form of "sunstroke syndrome + Guizhi Decoction syndrome") were extracted from the original text of Treatise on Febrile Diseases, and common synonyms that appear in these national standards, works, and modern Chinese medicine clinical practice were included. For specific contents, see Tables 10 and 11.

[0129] Table 10 Standardized table of disease names in Treatise on Febrile Diseases

[0130]

[0131] Table 11 Standard table of syndrome names in Treatise on Febrile Diseases (partial)

[0132]

[0133] The standardization of symptom names is based on the "Standardization of Common Clinical Symptom Terminology in Traditional Chinese Medicine (Revised)" to set the preferred terms for symptoms. Only when the symptom names are not mentioned in this book, the terms in the national higher education planning textbook for the traditional Chinese medicine industry "Selected Readings on the Treatise on Febrile Diseases (New Century 4th Edition)" shall prevail.

[0134] The names of treatments and therapies are standardized based on the classification structure and naming method of treatments and therapies in Selected Readings of Treatise on Febrile Diseases (New Century 4th Edition) and Annotated Treatise on Febrile Diseases. For the specific framework and naming standards, see Figure 5 .

[0135] The prescription names are standardized based on the prescription names in the original text of Treatise on Febrile Diseases. At the same time, according to the classification method of New Compilation of Treatise on Febrile Diseases by Mr. Liu Duzhou, all the prescriptions are divided into 18 categories, and two attributes, "indications" and "related articles", are set for each prescription entity. For the prescription classification method and name standard, please refer to Table 3.

[0136] In one of the embodiments, the predetermined technical specification is the TCMLS-SN technical specification; step 102 includes: according to the concept relationship description framework of Treatise on Febrile Diseases, referring to the TCMLS-SN technical specification, extracting knowledge from the original text of Treatise on Febrile Diseases by means of manual annotation + machine extraction + expert review, and constructing a concept hierarchy structure tree according to the hierarchical relationship.

[0137] In one embodiment, the concept nodes of the concept hierarchy structure tree include: Yin-Yang theory, six meridians, spirit, seven emotions, etiology and pathogenesis, disease, syndrome, symptom, treatment principle, treatment method, therapy, food, prescription, Chinese medicine, acupoints, meridians, viscera, basic substances of the human body, body organs, constitution, examination, examination results, people, geographical names, and Chinese medicine medical records; semantic relations include: upper and lower, equivalent, located, included, composed of, occurred before, occurred at the same time as, exterior and interior, use, phenomenon expression, main symptom, concurrent symptom, pulse symptom, necessary symptom, treatment, concurrent, caused, affected, misdiagnosis, forbidden, transmission, meaning extension, is...original disease, is...treatment history, is...current disease, is...diagnosis result, is...treatment plan; constraint rules are represented by triples consisting of head concept, semantic relationship, and tail concept, and the concept hierarchy structure tree includes 44 constraint rules. The constraint rules are shown in Table 12.

[0138] Table 12 Constraint rules of the knowledge system of Treatise on Febrile Diseases

[0139]

[0140]

[0141]

[0142] In one of the embodiments, the method further includes: receiving a query instruction, and returning reply information corresponding to the query instruction, wherein the query instruction includes a query on the knowledge structure of Treatise on Febrile Diseases, a query on the relevance of specific knowledge of Treatise on Febrile Diseases, and a query on medical records of Treatise on Febrile Diseases.

[0143] Specifically, by embedding the Cypher query statements provided by Neo4j in the program, we can obtain the sum of relationships centered on disease, syndrome, symptom, method, and prescription, realize the knowledge structure query of Treatise on Febrile Diseases, and display it; we can obtain the relationship network between specific knowledge in the "disease-syndrome-symptom-method-prescription" knowledge system of Treatise on Febrile Diseases, realize the correlation query of specific knowledge in Treatise on Febrile Diseases; and we can also perform medical record query. Fig.10 The conceptual hierarchy of Taiyang disease syndrome shown is the knowledge query result of Taiyang disease syndrome. Fig.11 The knowledge graph centered on the symptom "aversion to cold" is the specific knowledge association query result of "aversion to cold". Fig.12 The knowledge graph of medical records of mild syndrome of superficial depression in Treatise on Febrile Diseases shown is the result of medical record search related to mild syndrome of superficial depression.

[0144] In one embodiment, based on the semantics and relationships established by the concept hierarchy structure tree, as well as data preprocessing, knowledge extraction and knowledge fusion, the "Disease-Syndrome-Symptom-Method-Prescription" knowledge graph of "Treatise on Febrile Diseases" was constructed for the 10 articles in the original "Treatise on Febrile Diseases" involving disease identification, syndrome differentiation and treatment and 328 medical records of famous typhoid fever experts, including 6469 entities and 10911 relationship triples. Tables 13 and 14 show the statistical results of entities and relationships stored in Neo4j, respectively.

[0145] Table 13 Statistics of various entities stored in Neo4j graph database

[0146] Entity Type Chinese meaning quantity Entity Type Chinese meaning quantity Disease TCM disease 15 Formula prescription 112 Pattern Symptoms 299 TCM Case Chinese Medicine Cases 604 Symptom symptom 561 Others other 4866 Therapeutic methods Governing Law 12

[0147] Table 14 Statistical results of Neo4j graph database storage relationship triples

[0148]

[0149] It should be understood that although Figure 1 , 7 The steps in the flowchart of 8 are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 , 7 At least part of the steps in 8 may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0150] In one embodiment, Fig.13 As shown, a knowledge graph construction device based on Treatise on Febrile Diseases is provided, including: a concept relationship description framework construction module of Treatise on Febrile Diseases, a pattern layer construction module of the knowledge graph of Treatise on Febrile Diseases, and a data layer construction module of the knowledge graph of Treatise on Febrile Diseases, wherein:

[0151] The module for constructing the conceptual relationship description framework of Treatise on Febrile Diseases is used to analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, determine the relationship between disease, syndrome, symptom, method and prescription based on the conceptual system, and construct the conceptual relationship description framework of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription.

[0152] The pattern layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to extract knowledge from the original text of Treatise on Febrile Diseases according to the concept relationship description framework of Treatise on Febrile Diseases and reference predetermined technical specifications, and to construct a concept hierarchy structure tree according to the hierarchical relationship; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships.

[0153] The data layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to perform entity extraction, entity attribute extraction and entity relationship extraction on the original text of Treatise on Febrile Diseases according to concept nodes, semantic relationships and constraint rules to obtain a set of knowledge triples; perform knowledge fusion on the set of knowledge triples to obtain a set of knowledge triples after knowledge fusion and a synonym library of Treatise on Febrile Diseases; store the set of knowledge triples in the Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases.

[0154] In one of the embodiments, the device also includes a knowledge graph updating module based on the medical records of famous doctors of typhoid fever, which is used to perform entity extraction, entity attribute extraction and entity relationship extraction on the medical records of famous doctors of typhoid fever according to concept nodes, semantic relationships and constraint rules to obtain knowledge triples of the medical records of famous doctors of typhoid fever; map the entity names in the knowledge triples of the medical records of famous doctors of typhoid fever to standard entities according to the synonym library, and use statistical methods to calculate the entity similarity; the entity similarity is the weighted sum of character similarity and semantic similarity; the entity with the highest entity similarity and greater than a preset threshold is submitted as a candidate alignment entity to a Chinese medicine expert for review; if the review is passed, the attributes and relationships of the two entities are merged, and a synonym relationship is established between the entity and the standard entity in the disease-syndrome-symptom-method-prescription knowledge graph of Treatise on Febrile Diseases; if the review is not passed, the entity is added as a new entity to the disease-syndrome-symptom-method-prescription knowledge graph of Treatise on Febrile Diseases.

[0155] In one embodiment, the character similarity in the knowledge graph update module based on the medical records of famous typhoid fever is calculated by using the Jaccard distance between two entity strings, and the character similarity is calculated using formula (2); the semantic similarity is calculated by using the cosine similarity between two entities, and the semantic similarity is calculated using formula (3).

[0156] In one of the embodiments, the module for constructing the conceptual relationship description framework of Treatise on Febrile Diseases is also used to analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual classification system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, and use the relevant vocabulary collected from the five aspects of disease, syndrome, symptom, method and prescription as standard terms and add them into the term dictionary; according to the conceptual classification system of disease, syndrome, symptom, method and prescription, the relationship between disease, syndrome, symptom, method and prescription is defined by using conceptual relationship and attribute relationship, and the relationship between disease, syndrome, symptom, method and prescription includes: disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship and syndrome-method relationship. Relationship, syndrome-prescription relationship, method-prescription relationship; concept relationship is represented by the form of triple <concept A, relationship R, concept B>, where the types of relationship R include: one-to-one, one-to-many, many-to-many; attribute relationship is represented by the form of triple <concept A, attribute B, attribute value C>, where the type of attribute value is number or string, and the attribute value includes: name, duration, severity, location, nature of the disease; According to the disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship, syndrome-method relationship, syndrome-prescription relationship, method-prescription relationship, a conceptual relationship description framework of Treatise on Febrile Diseases is constructed.

[0157] In one embodiment, according to the concept nodes, semantic relationships and constraint rules, the data layer construction module of the Shanghan Lun knowledge graph also includes

[0158] The data preprocessing module is used to convert the original text of Treatise on Febrile Diseases into knowledge form under the guidance of TCM experts, based on concept nodes, semantic relationships and constraint rules, and with reference to pre-selected existing compilations of Treatise on Febrile Diseases, to obtain multiple knowledge tables; each cell content in the knowledge table has a clear scope, and synonyms and symptom types are also marked with special characters.

[0159] The entity extraction module is used to implement Chinese word segmentation in each cell of the knowledge table with the help of special characters and the NLPIR-ICTCLAS Chinese word segmentation system to complete the entity extraction of five aspects: disease, syndrome, symptom, method, and prescription.

[0160] The entity relationship extraction module is used to assign a specific entity type to each cell of the knowledge table and a specific relationship type between cells according to semantic relationships and constraint rules, and obtain knowledge table rules; based on the terminology dictionary and knowledge table rules, a set of knowledge triples is extracted from all knowledge tables to complete the entity relationship extraction of five aspects: disease, syndrome, symptom, method, and prescription.

[0161] In one of the embodiments, the data preprocessing module is also used for all the scriptures in Treatise on Febrile Diseases involving disease differentiation and syndrome differentiation. When each scripture involves only one diagnostic conclusion, knowledge is extracted from the original disease syndrome, treatment history, current disease syndrome, etiology and pathogenesis, syndrome differentiation results, treatment plan and plan coding to obtain a knowledge table corresponding to the text of Treatise on Febrile Diseases; each term in the knowledge type of current disease syndrome and treatment plan in the knowledge table indicates the concept node to which it belongs; when each scripture involves multiple diagnostic conclusions, the scripture is split into multiple scriptures so that the split scriptures contain only one diagnostic conclusion, and the knowledge extraction method involving only one diagnostic conclusion is used to extract the knowledge table corresponding to the text of Treatise on Febrile Diseases; the disease entities, syndrome entities, symptom entities, treatment method entities, therapy entities and prescription entities involved in the knowledge extraction process are named in accordance with the corresponding preset criteria.

[0162] In one of the embodiments, the predetermined technical specification is the TCMLS-SN technical specification; the pattern layer construction module of the Treatise on Febrile Diseases knowledge graph is also used to extract knowledge from the original text of the Treatise on Febrile Diseases by manual annotation + machine extraction + expert review based on the concept relationship description framework of the Treatise on Febrile Diseases, with reference to the TCMLS-SN technical specification, and to construct a concept hierarchy structure tree according to the hierarchical relationship.

[0163] In one embodiment, the concept nodes of the concept hierarchy structure tree in the device include: Yin-Yang theory, six meridians, spirit, seven emotions, etiology and pathogenesis, disease, syndrome, symptom, treatment principle, treatment method, therapy, food, prescription, Chinese medicine, acupoints, meridians, internal organs, basic substances of the human body, body organs, constitution, examination, examination results, people, geographical names, and Chinese medicine medical records; semantic relationships include: hierarchical, equivalent, located, containing, composed of, occurring before, occurring at the same time as, exterior and interior, use, phenomenon expression, main symptom, concomitant symptom, pulse symptom, necessary symptom, treatment, concurrent, causing, affecting, misdiagnosis, forbidden, transmission, meaning extension, is...original disease, is...treatment history, is...current disease, is...diagnosis result, is...treatment plan; constraint rules are represented by triples consisting of head concept, semantic relationship, and tail concept, and the concept hierarchy structure tree includes 44 constraint rules.

[0164] In one of the embodiments, the device further includes: receiving a query instruction, and returning reply information corresponding to the query instruction, wherein the query instruction includes a query on the knowledge structure of Treatise on Febrile Diseases, a query on the relevance of specific knowledge of Treatise on Febrile Diseases, and a query on medical records of Treatise on Febrile Diseases.

[0165] For the specific limitations of the knowledge graph construction device based on Treatise on Febrile Diseases, please refer to the limitations of the knowledge graph construction method based on Treatise on Febrile Diseases above, which will not be repeated here. Each module in the above-mentioned knowledge graph construction device based on Treatise on Febrile Diseases can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0166] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0167] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for constructing a knowledge graph based on Treatise on Febrile Diseases, characterized in that: The method comprises: Analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, determine the relationship between disease, syndrome, symptom, method and prescription based on the conceptual system, and construct a description framework of the conceptual relationship of Treatise on Febrile Diseases based on the relationship between disease, syndrome, symptom, method and prescription; According to the concept relationship description framework of Treatise on Febrile Diseases, with reference to predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a concept hierarchy structure tree is constructed according to the hierarchical relationship; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships; According to the concept nodes, the semantic relationships and the constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the original text of Treatise on Febrile Diseases to obtain a knowledge triple set; Performing knowledge fusion on the knowledge triple set to obtain a knowledge triple set after knowledge fusion and a synonym database of Treatise on Febrile Diseases; The knowledge triple set is stored in the Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases.

2. The method according to claim 1, characterized in that The method further comprises: According to the concept nodes, the semantic relationships and the constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the medical records of famous doctors of typhoid fever to obtain knowledge triples of the medical records of famous doctors of typhoid fever; According to the synonym library, the entity names in the knowledge triples of the famous medical records of typhoid fever are mapped to standard entities, and the entity similarity is calculated by a statistical method; the entity similarity is a weighted sum of character similarity and semantic similarity; The entity with the highest entity similarity greater than the preset threshold is submitted as a candidate alignment entity to TCM experts for review; if the review is passed, the attributes and relationships of the two entities are merged, and a synonym relationship is established between the entity and the standard entity in the disease-syndrome-symptom-method-prescription knowledge graph of Treatise on Febrile Diseases; if the review is not passed, the entity is added as a new entity to the disease-syndrome-symptom-method-prescription knowledge graph of Treatise on Febrile Diseases.

3. The method according to claim 2, characterized in that The character similarity is calculated using the Jaccard distance between two entity strings. The expression for character similarity is: Among them, J(A,B) is the character similarity between entity A and entity B, || is the number of characters to be found, |A∩B| is the character intersection of entity A and entity B, and |A∪B| is the character union of entity A and entity B; The semantic similarity is calculated using the cosine similarity between two entities. The expression of semantic similarity is: Among them, sim s (A,B) is the semantic similarity between entity A and entity B, (x1,x2,…,x n ) is the vector representation of entity A, (y1,y2,…,y n ) is the vector representation of entity B.

4. The method according to claim 1, characterized in that: The original text of Treatise on Febrile Diseases was analyzed and mined, and a conceptual system of disease, syndrome, symptom, method and prescription was established from the three aspects of theory, method and prescription. The relationship between disease, syndrome, symptom, method and prescription was determined based on the conceptual system, and a conceptual relationship description framework of Treatise on Febrile Diseases was constructed based on the relationship between disease, syndrome, symptom, method and prescription, including: Analyze and mine the original text of Treatise on Febrile Diseases, establish a conceptual classification system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, and collect relevant words from the five aspects of disease, syndrome, symptom, method and prescription as standard terms and add them into the term dictionary; According to the conceptual classification system of disease, syndrome, symptom, method and prescription, the relationship between disease, syndrome, symptom, method and prescription is defined by using conceptual relationship and attribute relationship, and the relationship between disease, syndrome, symptom, method and prescription includes: disease-disease relationship, disease-syndrome relationship, disease-symptom relationship, syndrome-symptom relationship, syndrome-method relationship, syndrome-prescription relationship, method-prescription relationship; the conceptual relationship is represented by the form of triple <concept A, relationship R, concept B>, where the types of relationship R include: one-to-one, one-to-many, many-to-many; the attribute relationship is represented by the form of triple <concept A, attribute B, attribute value C>, where the type of attribute value is number or string, and the attribute values ​​include: name, duration, severity, disease location, and disease nature; Based on the relationship between disease and disease, disease and syndrome, disease and symptom, syndrome and symptom, syndrome and method, syndrome and prescription, and method and prescription, a description framework of the conceptual relationship of Treatise on Febrile Diseases is constructed.

5. The method according to claim 4, characterized in that According to the concept nodes, the semantic relationships and the constraint rules, entity extraction, entity attribute extraction and entity relationship extraction are performed on the original text of Treatise on Febrile Diseases to obtain a knowledge triple set, including: Under the guidance of TCM experts, based on the concept nodes, the semantic relationships and the constraint rules, and with reference to the pre-selected existing compilations of Treatise on Febrile Diseases, the original text of Treatise on Febrile Diseases is converted into knowledge forms to obtain multiple knowledge tables; each cell content in the knowledge table has a clear range, and synonyms and symptom types are also marked with special characters; In each cell of the knowledge table, by means of the special characters, the NLPIR-ICTCLAS Chinese word segmentation system is used to implement Chinese word segmentation, and complete entity extraction in five aspects: disease, syndrome, symptom, method, and prescription; According to the semantic relationship and the constraint rules, a specific entity type is assigned to each cell of the knowledge table, as well as a specific relationship type between cells, to obtain a knowledge table rule; Based on the term dictionary and the knowledge table rules, a set of knowledge triples is extracted from all the knowledge tables to complete the extraction of entity relationships in five aspects: disease, syndrome, symptom, method and prescription.

6. The method according to claim 5, characterized in that Under the guidance of TCM experts, based on the concept nodes, semantic relationships and constraint rules, and with reference to the pre-selected existing compilations of Treatise on Febrile Diseases, the original text of Treatise on Febrile Diseases was converted into knowledge forms to obtain multiple knowledge tables, including: For all the texts in Treatise on Febrile Diseases involving disease and syndrome differentiation, when each text involves only one diagnostic conclusion, knowledge is extracted from the original disease and syndrome, treatment history, current disease and syndrome, etiology and pathogenesis, syndrome differentiation results, treatment plan and scheme coding to obtain the knowledge table corresponding to the text in Treatise on Febrile Diseases; each term in the knowledge type of current disease and syndrome and treatment plan in the knowledge table indicates the concept node to which it belongs; When each scripture involves multiple diagnostic conclusions, the scripture is split into multiple scriptures, so that the split scriptures only contain one diagnostic conclusion, and the knowledge extraction method involving only one diagnostic conclusion is used to extract the knowledge table corresponding to the article of Treatise on Febrile Diseases; The disease entities, syndrome entities, symptom entities, treatment method entities, therapy entities and prescription entities involved in the knowledge extraction process are named in accordance with the corresponding preset criteria.

7. The method according to claim 1, characterized in that The predetermined technical specification is the TCMLS-SN technical specification; According to the concept relationship description framework of Treatise on Febrile Diseases, and referring to the predetermined technical specifications, knowledge is extracted from the original text of Treatise on Febrile Diseases, and a concept hierarchy structure tree is constructed according to the hierarchical relationship, including: According to the concept relationship description framework of Treatise on Febrile Diseases, and referring to the technical specifications of TCMLS-SN, the knowledge of the original text of Treatise on Febrile Diseases was extracted by manual annotation + machine extraction + expert review, and a concept hierarchy structure tree was constructed according to the hierarchical relationship.

8. The method according to claim 7, characterized in that The concept nodes of the concept hierarchy structure tree include: Yin-Yang theory, six meridians, spirit, seven emotions, etiology and pathogenesis, disease, syndrome, symptom, treatment principle, treatment method, therapy, food, prescription, Chinese medicine, acupoints, meridians, internal organs, basic substances of the human body, body organs, constitution, examination, examination results, people, geographical names, and Chinese medical records; The semantic relations include: hyponymy, equivalence, located in, including, composed of, occurring before, occurring at the same time as, exterior and interior, use, phenomenon expression, main symptom, concurrent symptom, pulse symptom, necessary symptom, treatment, concurrent, cause, influence, misdiagnosis, forbidden, transmission, meaning extension, is the original disease, is the history of treatment, is the current disease, is the result of syndrome differentiation, is the treatment plan; The constraint rules are represented by a triple consisting of a head concept, a semantic relationship, and a tail concept. The concept hierarchy structure tree includes 44 constraint rules.

9. The method according to any one of claims 1 to 8, characterized in that: The method further comprises: receiving a query instruction, and returning reply information corresponding to the query instruction, wherein the query instruction comprises a query on the knowledge structure of Treatise on Febrile Diseases, a query on the relevance of specific knowledge of Treatise on Febrile Diseases, and a query on medical records of Treatise on Febrile Diseases.

10. A knowledge graph construction device based on Treatise on Febrile Diseases, characterized in that: The device comprises: The module for constructing the concept relationship description framework of Treatise on Febrile Diseases is used to analyze and mine the original text of Treatise on Febrile Diseases, establish the concept system of disease, syndrome, symptom, method and prescription from the three aspects of theory, method and prescription, determine the relationship between disease, syndrome, symptom, method and prescription according to the concept system, and construct the concept relationship description framework of Treatise on Febrile Diseases according to the relationship between disease, syndrome, symptom, method and prescription; The model layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to extract knowledge from the original text of Treatise on Febrile Diseases according to the concept relationship description framework of Treatise on Febrile Diseases and reference the predetermined technical specifications, and to construct a concept hierarchy structure tree according to the upper and lower relationships; the concept hierarchy structure tree includes concept nodes, semantic relationships, and constraint rules between concept nodes and semantic relationships; The data layer construction module of the knowledge graph of Treatise on Febrile Diseases is used to perform entity extraction, entity attribute extraction and entity relationship extraction on the original text of Treatise on Febrile Diseases according to the concept nodes, the semantic relationships and the constraint rules to obtain a set of knowledge triples; perform knowledge fusion on the set of knowledge triples to obtain a set of knowledge triples after knowledge fusion and a synonym library of Treatise on Febrile Diseases; store the set of knowledge triples using the Neo4j graph database to complete the construction of the knowledge graph of Treatise on Febrile Diseases.

Images