Traditional Chinese medicine knowledge graph construction method and system based on AI semantic understanding

Abstract

According to the traditional Chinese medicine knowledge graph construction method and system based on AI semantic understanding provided by the invention, a traditional Chinese medicine knowledge architecture is efficiently extracted and constructed from traditional Chinese medicine classical literature data, and systematic arrangement of traditional Chinese medicine theoretical knowledge is realized. And further, through deep correlation analysis of each knowledge concept entity in the traditional Chinese medicine knowledge architecture, not only is the logical relationship among the knowledge entities clear, but also the specific traditional Chinese medicine knowledge attribute characteristics of each correlated knowledge entity are described in detail, so that the content of the traditional Chinese medicine knowledge graph is greatly enriched. Finally, through the generation of the data stream which is represented in a mapping mode and displayed in a visualized mode, complex traditional Chinese medicine knowledge becomes visual and easy to understand, data support is provided for inheritance, research and clinical application of the traditional Chinese medicine knowledge, and accessibility and practicability of the traditional Chinese medicine knowledge are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for constructing a traditional Chinese medicine knowledge graph based on AI semantic understanding. Background Art

[0002] Traditional Chinese medicine carries thousands of years of medical wisdom and practical experience. However, with the development of the times, the inheritance and popularization of traditional Chinese medicine knowledge faces unprecedented challenges. The traditional way of learning traditional Chinese medicine mainly relies on master-apprentice teaching and reading of classic literature. Although this method can pass on the essence of traditional Chinese medicine in depth, it is inefficient and is limited by factors such as region and time, making it difficult to be widely popularized.

[0003] In recent years, with the rapid development of information technology, especially the rise of artificial intelligence (AI) technology, it has provided a new opportunity for the inheritance and popularization of TCM knowledge. AI technology can simulate human intelligence and efficiently process and analyze large amounts of data, which provides strong support for the systematic and structured organization of TCM knowledge. However, most of the TCM knowledge systems on the market currently remain at the simple information retrieval level, lacking a deep semantic understanding of TCM theory and knowledge association analysis.

[0004] As an effective way of knowledge representation, TCM knowledge graph can intuitively display the knowledge concept entities and their interrelationships in TCM theory in the form of graphs, which helps users better understand and master TCM knowledge. However, how to construct an accurate and comprehensive TCM knowledge graph has always been an urgent problem to be solved in the field of TCM informatization. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for constructing a TCM knowledge graph based on AI semantic understanding, the method comprising: Acquire TCM classic literature data, and construct a TCM knowledge framework from the TCM classic literature data; the TCM classic literature data is a data source corresponding to TCM theory, the TCM knowledge framework is a feature vector of a TCM knowledge graph to be constructed, and the TCM knowledge framework includes one or more knowledge concept entities; Performing association analysis on each knowledge concept entity in the TCM knowledge architecture, generating one or more associated knowledge entities corresponding to each knowledge concept entity in the TCM knowledge graph, and semantic relationship information and knowledge feature vector information corresponding to each associated knowledge entity in the one or more associated knowledge entities, wherein the semantic relationship information is the logical relationship between each associated knowledge entity and the knowledge concept entity, and the knowledge feature vector information is the specific TCM knowledge attribute feature represented by each associated knowledge entity; Wandering the TCM knowledge graph, based on the semantic relationship information and the knowledge feature vector information, graphically represent each associated knowledge entity, and generate a visual display data stream of the TCM knowledge graph.

[0006] On the other hand, an embodiment of the present invention also provides a system for constructing a traditional Chinese medicine knowledge graph based on AI semantic understanding, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0007] Based on the above aspects, the embodiment of the present application realizes the systematic arrangement of TCM theoretical knowledge by efficiently extracting and constructing a TCM knowledge framework from TCM classic literature data. Furthermore, through the in-depth association analysis of each knowledge concept entity in the TCM knowledge framework, not only the logical relationship between each knowledge entity is clarified, but also the specific TCM knowledge attribute characteristics of each associated knowledge entity are described in detail, thereby greatly enriching the content of the TCM knowledge map. Finally, through the generation of graphical representation and visual display data flow, complex TCM knowledge becomes intuitive and easy to understand, providing data support for the inheritance, research and clinical application of TCM knowledge, and significantly improving the accessibility and practicality of TCM knowledge. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic diagram of the execution flow of the method for constructing a traditional Chinese medicine knowledge graph based on AI semantic understanding provided in an embodiment of the present invention.

[0009] Figure 2 It is a schematic diagram of the hardware architecture of a TCM knowledge graph construction system based on AI semantic understanding provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 It is a flow chart of a method for constructing a TCM knowledge graph based on AI semantic understanding provided by an embodiment of the present invention. The method for constructing a TCM knowledge graph based on AI semantic understanding is introduced in detail below.

[0011] Step S110, obtaining TCM classic literature data, and constructing a TCM knowledge framework from the TCM classic literature data. The TCM classic literature data is a data source corresponding to TCM theory, the TCM knowledge framework is a feature vector of a TCM knowledge graph to be constructed, and the TCM knowledge framework includes one or more knowledge concept entities.

[0012] In this embodiment, the server first obtains the data of TCM classic literature. For a certain TCM research institution, the server is connected to a database storing a large number of TCM related materials. The TCM classic literature data in this database comes from a wide range of sources, which can be the content of the digitized manuscripts, or the literature on traditional TCM diagnosis and treatment experience collected and sorted by the TCM R&D team over many years, which contains many unique TCM theoretical insights.

[0013] For example, there is a manuscript that records the ideas of treating special diseases passed down in a mountainous area, which involves the understanding of the changes of human qi and blood under specific circumstances. This understanding is somewhat different from the common Chinese medicine theory of qi and blood. The server will also include this type of data in the acquisition scope.

[0014] After the server obtains these TCM classic literature data, it starts to build the TCM knowledge architecture. Suppose the server obtains a document about a certain regional TCM theory, which mentions a special concept of "qi". This concept has a unique connection with the five internal organs of the human body, which is different from the general understanding of the relationship between qi and internal organs in traditional TCM theory. The server will extract this special "qi" concept as a knowledge concept entity. At the same time, another document mentions the relationship between a special herbal preparation method and the human meridians. The server will also extract related elements such as the herbal preparation method, the human meridians, and the relationship between them as knowledge concept entities. With the analysis of more literature data, the server gradually builds the TCM knowledge architecture. The knowledge concept entities in this architecture are intertwined to form a complex network, just like the nodes in a large network. Each node represents a knowledge concept entity. These knowledge concept entities cover various aspects from special diseases, unique treatment methods to unique TCM theoretical concepts. Together, they constitute the feature vectors of the TCM knowledge graph to be constructed.

[0015] Step S120, performing association analysis on each knowledge concept entity in the TCM knowledge architecture, generating one or more associated knowledge entities corresponding to each knowledge concept entity in the TCM knowledge graph, and semantic relationship information and knowledge feature vector information corresponding to each associated knowledge entity in the one or more associated knowledge entities, wherein the semantic relationship information is the logical relationship between each associated knowledge entity and the knowledge concept entity, and the knowledge feature vector information is the specific TCM knowledge attribute characteristics represented by each associated knowledge entity.

[0016] In this embodiment, the server begins to perform association analysis on each knowledge concept entity in the constructed TCM knowledge architecture. Take the previously mentioned special "qi" concept knowledge concept entity as an example. The server will first analyze the overall structure of the TCM knowledge architecture, in which the special concept of "qi" has various potential connections with other knowledge concept entities. From the overall structure, the area where the concept of "qi" is located may have an indirect connection with the area of ​​some special diseases, because some documents mention that changes in this special "qi" may cause some special diseases. At the same time, there may also be a connection between the concept of "qi" and some special acupuncture points of the human body. Although this connection is not clearly stated in traditional TCM theory, there are relevant hints in these unique documents.

[0017] Then, the server determines the initial semantic domain scope involved in the conceptual knowledge concept entity of "qi". This scope may cover aspects such as human physiological functions and the inducing factors of special diseases. Then, the server retrieves and screens the data of classical Chinese medicine literature based on this initial semantic domain scope. During the screening process, a collection of text descriptions related to the initial semantic domain scope will be mined. For example, in a document on the treatment of local special diseases, there is a description like this: "When this qi goes backwards, the heart pulse is not smooth, and this symptom is seen in those with dull pain between the eyebrows." This description is included in the text description collection.

[0018] When parsing this text description set, the server analyzes the basic logical relationship between the grammatical structural elements in the sentence, as well as the TCM connotation of each word and phrase. In the above sentence, there is a causal relationship between "retrograde qi" and "poor heart pulse", and there is also a causal relationship between "poor heart pulse" and "dull pain between eyebrows", which are the basic logical relationships between the grammatical structural elements. From the TCM connotation, "retrograde qi" here means that the flow direction of this special "qi" is opposite to the normal situation, "poor heart pulse" means that the flow of qi and blood in the meridians related to the heart is blocked, and "dull pain between eyebrows" is a symptom. Based on these relationships and connotations, the server identifies potential related elements related to the concept of "qi", such as "heart pulse", "eyebrow", "retrograde qi", etc., and generates a list of potential related elements.

[0019] According to the logic and classification standards of the TCM knowledge system, the server classifies this list of potential related elements. For the element "heart pulse", it belongs to the category of human meridians; "eyebrow center" belongs to the category of human body parts; "retrograde qi" belongs to the category of special qi states. Then, for each classification category, the degree of association between the element and the knowledge concept entity "qi" is evaluated. When evaluating the degree of association between "heart pulse" and "qi", based on the causal relationship in TCM theory (because retrograde qi can cause poor heart pulse), it can be judged that they are closely related and conform to the logic of TCM knowledge. For "eyebrow center", since it is the site of symptom manifestation after poor heart pulse caused by retrograde qi, it also has a certain association with "qi". After screening, these elements are considered to be closely related and conform to the logic of TCM knowledge, thereby generating a subset of valid related elements.

[0020] Each valid associated element in the valid associated element subset is determined as an associated knowledge entity. For example, "heart pulse", "eyebrow center", and "retrograde qi" have all become associated knowledge entities associated with the conceptual knowledge entity of "qi". According to the basic logical relationship between the grammatical structure component elements in the text description set, the semantic relationship information between each associated knowledge entity and the knowledge concept entity is constructed. The semantic relationship information between "heart pulse" and "qi" is "retrograde qi can cause poor heart pulse", which is a causal relationship; the semantic relationship information between "eyebrow center" and "qi" is "retrograde qi causes poor heart pulse and then causes dull pain in the eyebrow center", which is an indirect causal relationship; the semantic relationship information between "retrograde qi" and "qi" itself is "it is a special state of qi". At the same time, for each associated knowledge entity, the knowledge feature vector information of the associated knowledge entity is extracted according to the attribute feature definition in the traditional Chinese medicine knowledge system. For the "heart meridian", its knowledge feature vector information may include the meridian system to which it belongs, the characteristics of the circulation of qi and blood, etc.; for the "eyebrow center", its knowledge feature vector information may include its position characteristics on the human head, its relationship with the surrounding meridian acupoints, etc.; for the "retrograde flow of qi", its knowledge feature vector information may include the factors that cause its retrograde flow, the impact on other aspects of the human body during retrograde flow, etc.

[0021] For other knowledge concept entities in the TCM knowledge architecture, such as the herbal preparation method knowledge concept entity mentioned above, the server will also perform similar association analysis. Assume that this herbal preparation method is a special combination of steaming and boiling, which has a special connection with the human meridians. By analyzing the overall structure and contextual relationship of the architecture, the server determines that its initial semantic domain scope may involve aspects such as the transformation of herbal medicinal properties and the effect on human meridians. By searching and screening the literature data, a text description set containing descriptions such as "this preparation method can make the medicinal properties of herbs enter the kidney meridian and open the lower meridians" is obtained. These sentences are parsed to identify potential associated elements such as "kidney meridian", "lower meridian", "herbal medicinal property transformation", etc., and after classifying and evaluating the degree of association, a subset of valid associated elements is obtained. The “Kidney Meridian”, “Lower Jiao Meridian” and “Transformation of Herbal Medicinal Properties” were identified as related knowledge entities, and semantic relationship information was constructed (such as “This processing method transforms the medicinal properties of herbs and then enters the kidney meridian and passes through the lower Jiao meridian”). The knowledge feature vector information of each related knowledge entity was extracted (such as the meridian direction of the “Kidney Meridian” and its connection with the internal organs, the scope and functional characteristics of the “Lower Jiao Meridian”, the transformation conditions of the “Transformation of Herbal Medicinal Properties”, and the characteristics of the medicinal properties after transformation, etc.).

[0022] Step S130, walk the TCM knowledge graph, and based on the semantic relationship information and the knowledge feature vector information, perform a graph representation of each associated knowledge entity to generate a visual display data stream of the TCM knowledge graph.

[0023] In this embodiment, the server starts to walk through the already constructed TCM knowledge graph. Assume that the walk starts from the special concept knowledge entity of "Qi" and its associated knowledge entities in the previously constructed TCM knowledge graph. First, based on the semantic relationship information in each associated knowledge entity, the associated knowledge concept entity is determined. Taking the associated knowledge entity "Qi Retrograde" as an example, according to its semantic relationship information with "Qi" "is a special state of Qi", it can be determined that its associated knowledge concept entity is "Qi". Then, the basic information of traditional Chinese medicine corresponding to the associated knowledge concept entity is obtained. For the knowledge concept entity of "Qi", its basic information of traditional Chinese medicine may include the normal operation law of Qi in the human body, the source of Qi, etc.

[0024] Based on the knowledge feature vector information and the traditional Chinese medicine basic information corresponding to the associated knowledge concept entity, the feature vector corresponding to the current display node in each associated knowledge entity is represented graphically to generate the current Chinese medicine knowledge display node. For the associated knowledge entity "Qi retrograde", combined with its knowledge feature vector information (factors causing its retrograde, the impact on other aspects of the human body during retrograde, etc.) and the traditional Chinese medicine basic information of "Qi" (the normal operation law of Qi in the human body, etc.), the relevant features of the "Qi retrograde" node can be represented in the graph. For example, in the graph, this node may be displayed with a specific color (indicating an abnormal state), a specific shape (indicating the relationship with Qi), and some text descriptions (such as factors causing retrograde, etc.), which generates the current Chinese medicine knowledge display node, which belongs to the visualization display data flow of the Chinese medicine knowledge graph.

[0025] Next, the server continues to walk through the TCM knowledge graph and graphically represents the feature vector corresponding to the next display node. For example, when walking to the associated knowledge entity "heart pulse", according to its semantic relationship information with "qi" "qi retrograde can cause heart pulse obstruction", it is determined that its associated knowledge concept entity is "qi", and the traditional TCM basic information of "qi" is obtained. Then, combined with the knowledge feature vector information of "heart pulse" (the meridian system to which it belongs, the characteristics of qi and blood circulation, etc.) and the traditional TCM basic information of "qi", the relevant features of the node "heart pulse" are represented in the graph, such as using different line colors to represent its qi and blood circulation status (due to qi retrograde leading to obstruction, lighter colors may be used to represent weak qi and blood circulation), using specific line thickness to represent its importance in the meridian system, etc., thus generating a new TCM knowledge display node. In this way, the server continuously walks through the TCM knowledge graph and graphically represents the feature vectors corresponding to each display node until the graphical representation of the feature vectors corresponding to each display node is completed, and the graphical representation is completed to generate a visual display data stream of the TCM knowledge graph. This visual display of data flow is like a dynamic TCM knowledge map, showing the complex relationships and characteristics between TCM knowledge concept entities and their related knowledge entities from different perspectives, providing an intuitive and comprehensive display of the TCM knowledge system for TCM researchers, practitioners, etc.

[0026] The same is true for the walk starting from the herbal preparation method knowledge concept entity and its associated knowledge entities. For example, starting from the associated knowledge entity "herbal medicine property transformation", according to its semantic relationship information with herbal preparation methods, the associated knowledge concept entity is determined to be herbal preparation methods. The traditional Chinese medicine basic information of herbal preparation methods (such as the purpose of preparation, traditional operation specifications, etc.) is obtained, combined with the knowledge feature vector information of "herbal medicine property transformation" (transformation conditions, characteristics of the transformed medicine property, etc.), and the characteristics of the node "herbal medicine property transformation" are represented in the graph, such as using a specific icon to represent the transformation process, using different colors to represent the category of the transformed medicine property, etc. Then, it walks to the associated knowledge entity "kidney meridian", and similarly determines the associated knowledge concept entity based on the semantic relationship information, obtains the relevant traditional Chinese medicine basic information, and combines the knowledge feature vector information of "kidney meridian" (meridian direction, connection with internal organs, etc.) to represent the characteristics of the node "kidney meridian" in the graph, such as using lines to represent the meridian direction, using node size to represent the closeness of the connection with internal organs, etc., and so on, and finally completes the generation of the visualization display data flow of the entire Chinese medicine knowledge graph.

[0027] Based on the above steps, the embodiment of the present application realizes the systematic arrangement of TCM theoretical knowledge by efficiently extracting and constructing TCM knowledge framework from TCM classic literature data. Furthermore, through the in-depth association analysis of each knowledge concept entity in the TCM knowledge framework, not only the logical relationship between each knowledge entity is clarified, but also the specific TCM knowledge attribute characteristics of each associated knowledge entity are described in detail, thereby greatly enriching the content of the TCM knowledge map. Finally, through the generation of graphical representation and visual display data flow, complex TCM knowledge becomes intuitive and easy to understand, providing data support for the inheritance, research and clinical application of TCM knowledge, and significantly improving the accessibility and practicality of TCM knowledge.

[0028] In a possible implementation, before step S130, the method further includes: Step A110, obtaining an adjustment instruction for the TCM knowledge graph. The adjustment instruction is instruction information for adjusting the TCM knowledge graph.

[0029] Step A120, adjusting the TCM knowledge graph based on the adjustment instruction to generate a target TCM knowledge graph.

[0030] Wherein, step S130 includes: The target TCM knowledge graph is walked, and each target-related knowledge entity in the target TCM knowledge graph is represented graphically based on the target semantic relationship information and the target knowledge feature vector information, so as to generate a visual display data stream of the TCM knowledge graph. The target semantic relationship information and the target knowledge feature vector information correspond to each target-related knowledge entity.

[0031] In this embodiment, in the server system of the TCM research institution, adjustment instructions for the TCM knowledge graph from different sources will be received. For example, a team of experts studying the TCM knowledge graph may send adjustment instructions to the server based on their research needs or new research results. This adjustment instruction is to adjust the constructed TCM knowledge graph to better reflect the accuracy of TCM knowledge or to be more in line with a specific research direction.

[0032] Taking the concept of the special "qi" mentioned earlier and the related knowledge graph as an example, if the expert team discovers new research results on the relationship between this special "qi" and other elements, they will send adjustment instructions. For example, if a special environmental factor is found to have an undiscovered effect on the running state of this "qi", and this effect involves a new human physiological reaction, this information will be organized into adjustment instructions and sent to the server.

[0033] After receiving this adjustment instruction, the server starts to adjust the TCM knowledge graph based on this instruction, thereby generating a target TCM knowledge graph. If the adjustment instruction is a category knowledge adjustment instruction, for example, it is to adjust the TCM knowledge attribute characteristics corresponding to the previously mentioned "Qi Retrograde" associated knowledge entity. The server will obtain the target associated knowledge entity "Qi Retrograde" and the knowledge feature vector information to be adjusted from this category knowledge adjustment instruction. This knowledge feature vector information to be adjusted may be about the new conditions or impact results of "Qi Retrograde" under the influence of this special environmental factor. Then the server obtains the target knowledge feature vector information corresponding to "Qi Retrograde" and updates the target knowledge feature vector information based on the knowledge feature vector information to be adjusted. For example, the original knowledge feature vector information about "Qi Retrograde" does not include the influence of special environmental factors. Now this new influencing factor and its related results are added to the knowledge feature vector information, so that the adjustment of this part of the TCM knowledge graph is completed and the target TCM knowledge graph is generated.

[0034] If the adjustment instruction is an architecture adjustment instruction, for example, an adjustment to the architecture of the TCM knowledge graph, this architecture adjustment instruction may include at least one of associated knowledge entity expansion information, associated knowledge entity destruction information, and associated knowledge entity migration information.

[0035] If the architecture adjustment instruction includes the associated knowledge entity extension information, the first target associated knowledge entity is obtained from the associated knowledge entity extension information, assuming that it is a group associated knowledge entity in the knowledge section about "qi", such as an associated knowledge entity group of a certain type of human physiological reaction related to the special "qi", and the associated knowledge entity to be extended, which may be the newly discovered special substance related to this physiological reaction in the human body. The server determines the knowledge concept entity to be extended based on the logical relationship between the associated knowledge entity to be extended and the knowledge concept entity, which may be the human physiological regulation concept entity related to the special "qi" concept. Then, the knowledge concept entity to be extended is analyzed for association, and the target associated knowledge entity to be extended is generated, such as a new associated knowledge entity in the human physiological regulation process related to the special substance reaction. Finally, the target associated knowledge entity to be extended is loaded into the first target associated knowledge entity of the traditional Chinese medicine knowledge graph. After the adjustment, if the category of the first target associated knowledge entity was originally a category associated knowledge entity, it may now become a group associated knowledge entity category, so that the architecture adjustment of the traditional Chinese medicine knowledge graph is ended and the target traditional Chinese medicine knowledge graph is generated.

[0036] If the architecture adjustment instruction includes the associated knowledge entity destruction information, the second target associated knowledge entity is obtained from the associated knowledge entity destruction information, assuming that it is a knowledge entity that was previously proven to be incorrectly associated with the concept of the special "qi", such as a knowledge entity associated with a non-existent physiological phenomenon that was mistakenly believed to be directly related to the special "qi". The server destroys the second target associated knowledge entity from the TCM knowledge graph, thereby ending the architecture adjustment of the TCM knowledge graph and generating the target TCM knowledge graph.

[0037] If the architecture adjustment instruction includes the associated knowledge entity migration information, the third target associated knowledge entity and the fourth target associated knowledge entity are obtained from the associated knowledge entity migration information, assuming that the third target associated knowledge entity is an associated knowledge entity with the special "qi" in a certain local theory, and the fourth target associated knowledge entity is an associated knowledge entity in another related but different theory, and they are determined by the knowledge section corresponding to the TCM knowledge graph, and there is no derivative relationship between the two associated knowledge entities. The server migrates these two associated knowledge entities in the TCM knowledge graph, such as adjusting their positions in the graph to make them more consistent with the logical relationship of TCM knowledge, ends the architecture adjustment of the TCM knowledge graph, and generates the target TCM knowledge graph.

[0038] After completing these adjustments, the server starts to roam the target TCM knowledge graph, and based on the target semantic relationship information and target knowledge feature vector information, it graphs each target-related knowledge entity in the target TCM knowledge graph, and generates a visual display data stream of the TCM knowledge graph.

[0039] Starting from the previously adjusted special "qi" related knowledge graph part, take the target associated knowledge entity "qi retrograde" as an example, based on its corresponding target semantic relationship information (such as the relationship with other elements after the new adjustment) and target knowledge feature vector information (including updated information such as new influencing factors), determine that the associated knowledge concept entity is "qi", obtain the traditional Chinese medicine basic information corresponding to "qi", and then combine this information to map the feature vector corresponding to the current display node in the target associated knowledge entity "qi retrograde", and generate the current Chinese medicine knowledge display node. This node belongs to the visualization display data flow of the Chinese medicine knowledge graph. Then continue to wander the target Chinese medicine knowledge graph in this way, and map the feature vector corresponding to the next display node until the map representation of the feature vector corresponding to each display node is completed, and the visualization display data flow of the Chinese medicine knowledge graph is generated. This process is like taking a tour in the newly constructed and adjusted Chinese medicine knowledge building, accurately displaying each knowledge entity according to the new rules and information, and forming an accurate, comprehensive, and visual display data flow of the Chinese medicine knowledge graph that meets the latest research results and needs.

[0040] In a possible implementation, the adjustment instruction is a category knowledge adjustment instruction. The category knowledge adjustment instruction is instruction information for adjusting the attribute characteristics of TCM knowledge corresponding to the associated knowledge entity in the TCM knowledge graph.

[0041] Step A120 includes: From the category knowledge adjustment instruction, the target associated knowledge entity and the feature vector information of the knowledge to be adjusted are obtained. The target associated knowledge entity is the associated knowledge entity in the traditional Chinese medicine knowledge graph whose knowledge feature is to be adjusted.

[0042] Obtain target knowledge feature vector information corresponding to the target associated knowledge entity.

[0043] Based on the knowledge feature vector information to be adjusted, the target knowledge feature vector information is updated, thereby completing the adjustment of the TCM knowledge graph and generating the target TCM knowledge graph.

[0044] In this embodiment, in a TCM research institution, the server is responsible for maintaining and adjusting the TCM knowledge graph. When the received adjustment instruction is a category knowledge adjustment instruction, it means that the TCM knowledge attribute features corresponding to the associated knowledge entities in the TCM knowledge graph need to be adjusted.

[0045] Taking the special "Qi"-related TCM knowledge graph mentioned earlier as an example, suppose that due to new research results or more in-depth excavation of ancient books, new information about the associated knowledge entity "Qi Retrograde" is discovered. At this time, the server obtains the target associated knowledge entity "Qi Retrograde" and the feature vector information of the knowledge to be adjusted from the category knowledge adjustment instruction. This feature vector information of the knowledge to be adjusted may come from new research findings. For example, it is found that under a certain special dietary structure, "Qi Retrograde" will have different manifestations and different effects on other functions of the human body. These new manifestations and effects are the feature vector information of the knowledge to be adjusted.

[0046] Then, the server obtains the target knowledge feature vector information corresponding to "Qi retrograde". This target knowledge feature vector information is the knowledge attribute feature about "Qi retrograde" that already exists in the current TCM knowledge graph, such as the previously recorded information about the impact of "Qi retrograde" on the circulation of Qi and blood in the human body and the functions of internal organs.

[0047] Next, based on the knowledge feature vector information to be adjusted, the target knowledge feature vector information is updated. The server integrates the newly discovered different manifestations and effects of "Qi Retrograde" under a special diet structure into the original target knowledge feature vector information. For example, the original target knowledge feature vector information only mentioned that "Qi Retrograde" can cause poor heart circulation, but now it has been added that under a special diet structure, "Qi Retrograde" will also affect the spleen and stomach transportation function. At the same time, new content has been added to the manifestation of "Qi Retrograde", such as the possibility of new manifestations such as abdominal distension. Through such an update, the adjustment of the knowledge feature vector information of the associated knowledge entity "Qi Retrograde" is completed, thereby completing the adjustment of this part of the TCM knowledge graph and generating the target TCM knowledge graph.

[0048] For another example, for the associated knowledge entity "herbal medicinal property transformation" related to the special herbal preparation method, if a category knowledge adjustment instruction is received. The target associated knowledge entity "herbal medicinal property transformation" and the knowledge feature vector information to be adjusted are obtained from the instruction. Suppose new research finds that different storage environments will have a special impact on the medicinal property transformation of herbs. The information related to this impact is the knowledge feature vector information to be adjusted. The server obtains the existing target knowledge feature vector information of "herbal medicinal property transformation", including the currently recorded transformation conditions, the medicinal property characteristics after transformation, etc. Then, based on the new knowledge feature vector information to be adjusted, the special impact of different storage environments on the medicinal property transformation of herbs is added to the target knowledge feature vector information, such as the special changes that may occur in the medicinal property transformation of herbs stored in a humid environment during preparation. After the adjustment is completed, the target Chinese medicine knowledge graph is generated.

[0049] This adjustment process based on category knowledge adjustment instructions enables the TCM knowledge graph to continuously absorb new knowledge content, more accurately reflect the connotation and extension of TCM knowledge, and provide more precise knowledge basis for TCM research, teaching, clinical practice and other aspects.

[0050] In a possible implementation, the adjustment instruction is an architecture adjustment instruction. The architecture adjustment instruction is instruction information for adjusting the architecture of the TCM knowledge graph.

[0051] Step A120 includes: Based on the architecture adjustment instruction, the architecture of the TCM knowledge graph is adjusted to generate the target TCM knowledge graph. The architecture adjustment instruction includes at least one of associated knowledge entity expansion information, associated knowledge entity destruction information, and associated knowledge entity migration information.

[0052] Among them, the associated knowledge entity extension information is the feature vector for expanding the associated knowledge entity into the traditional Chinese medicine knowledge graph, the associated knowledge entity destruction information is the feature vector for destroying the associated knowledge entity from the traditional Chinese medicine knowledge graph, and the associated knowledge entity migration information is the feature vector for migrating the associated knowledge entity in the traditional Chinese medicine knowledge graph.

[0053] In this embodiment, taking the special "qi" related to the TCM knowledge graph as an example, if the architecture adjustment instruction contains the associated knowledge entity extension information. Assume that in the further study of ancient TCM books, a new health-preserving method related to the special "qi" is discovered. This health-preserving method is closely related to the operation of the special "qi" in the human body, and is also associated with some human meridian knowledge entities in the previous graph. The server obtains the first target associated knowledge entity from the associated knowledge entity extension information, such as the health-preserving method knowledge section related to the special "qi" (this section can be regarded as a group associated knowledge entity), and the associated knowledge entity to be expanded, that is, the knowledge entity related to the newly discovered health-preserving method. Based on the logical relationship between the associated knowledge entity to be expanded and the knowledge concept entity (here is the special "qi"), the server determines the knowledge concept entity to be expanded, such as the human body spirit, qi and spirit concept entity that interacts with the special "qi" in the health-preserving method. Then, the association analysis is performed on this knowledge concept entity to be expanded to generate the target associated knowledge entity to be expanded, such as the associated knowledge entity related to the specific operation trajectory and changes of the spirit, qi and spirit in the new health-preserving method. Finally, the target-associated knowledge entity to be expanded is loaded into the first target-associated knowledge entity of the TCM knowledge graph. If the first target-associated knowledge entity is originally a category-associated knowledge entity, its category may become a group-associated knowledge entity category after loading, thereby completing the expansion and adjustment of this part of the architecture.

[0054] If the architecture adjustment instruction contains the destruction information of the associated knowledge entity, for example, in the knowledge graph related to the special "qi", if after more rigorous verification, it is found that the associated knowledge entity in the previous graph related to the special "qi" in the ancient legend (such as the associated knowledge entity of the special "qi" and the influence of a certain mythical creature on the human body) does not conform to the scientific logic of traditional Chinese medicine, the server obtains the second target associated knowledge entity from the destruction information of the associated knowledge entity, that is, the associated knowledge entity related to the ancient legend, and then destroys the second target associated knowledge entity from the traditional Chinese medicine knowledge graph, thereby adjusting this part of the architecture.

[0055] If the architecture adjustment instruction contains the migration information of related knowledge entities, for example, in the knowledge graph part related to the special "qi", there is an associated knowledge entity about the special "qi" in the local folk remedies, and another associated knowledge entity in the orthodox Chinese medicine theory. There is no derivative relationship between these two associated knowledge entities, but they are both in the corresponding knowledge section of the Chinese medicine knowledge graph. The server obtains the third target associated knowledge entity (the associated knowledge entity in folk remedies) and the fourth target associated knowledge entity (the associated knowledge entity in orthodox Chinese medicine theory) from the associated knowledge entity migration information, and then migrates these two associated knowledge entities in the Chinese medicine knowledge graph, such as migrating the associated knowledge entity in folk remedies to a more logical knowledge section related to folk medicine, and migrating the associated knowledge entity in orthodox Chinese medicine theory to a section that better reflects its theoretical core, so that the architecture of the Chinese medicine knowledge graph is more reasonable.

[0056] Similar operations are also performed for the structural adjustment of other parts of the TCM knowledge graph, such as herbal preparation methods. For example, for the knowledge graph related to herbal preparation methods, if new herbal preparation tools are found to be associated with existing preparation methods and herbal knowledge entities, the structural expansion can be performed through the associated knowledge entity expansion information; if an incorrect preparation-related knowledge entity is found in the previous graph (such as a knowledge entity that mistakenly believes that an irrelevant substance is involved in the preparation process), the structural adjustment can be performed through the associated knowledge entity destruction information; if there is an unreasonable position of the associated knowledge entity of the preparation method in different regional theories, it can be adjusted through the associated knowledge entity migration information. Through these operations, the adjustment of the TCM knowledge graph architecture is completed based on the architectural adjustment instructions, and the target TCM knowledge graph is generated, so that the architecture of the TCM knowledge graph reflects the TCM knowledge system more scientifically, reasonably, and accurately.

[0057] In a possible implementation, when the architecture adjustment instruction includes the associated knowledge entity extension information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: From the associated knowledge entity extension information, a first target associated knowledge entity and an associated knowledge entity to be extended are obtained, wherein the first target associated knowledge entity is a category associated knowledge entity or a group associated knowledge entity in the TCM knowledge graph for which feature vector extension is to be performed. The associated knowledge entity to be extended is an initial associated knowledge entity to be loaded into the first target associated knowledge entity.

[0058] Based on the logical relationship between the associated knowledge entity to be expanded and the knowledge concept entity, the knowledge concept entity to be expanded is determined.

[0059] Perform association analysis on the knowledge concept entity to be expanded to generate a target associated knowledge entity to be expanded.

[0060] The target associated knowledge entity to be expanded is loaded into the first target associated knowledge entity of the TCM knowledge graph, the structural adjustment of the TCM knowledge graph is completed, and the target TCM knowledge graph is generated, the first target associated knowledge entity and the target associated knowledge entity to be expanded are the same-level associated knowledge entities or the hierarchical associated knowledge entities, and when the first target associated knowledge entity is a category associated knowledge entity, and the first target associated knowledge entity and the target associated knowledge entity to be expanded are hierarchical associated knowledge entities, the category of the first target associated knowledge entity after adjustment is a group associated knowledge entity category.

[0061] In this embodiment, when the architecture adjustment instruction includes associated knowledge entity extension information, the server starts to perform a series of operations to adjust the architecture of the TCM knowledge graph to generate a target TCM knowledge graph.

[0062] Taking the special "qi"-related TCM knowledge graph as an example, suppose that after studying some newly unearthed ancient TCM book fragments, new knowledge content is found that needs to be expanded into the graph. The server obtains the first target associated knowledge entity and the associated knowledge entity to be expanded from the associated knowledge entity extension information. For example, the first target associated knowledge entity is the health-preserving exercise knowledge section related to the special "qi". This section is currently a category-associated knowledge entity in the graph, which contains some basic health-preserving exercises and the associated knowledge of "qi". The associated knowledge entity to be expanded is a knowledge entity related to the specific action steps of an ancient health-preserving exercise found in the newly unearthed ancient books and its special connection with the special "qi". This entity is the initial associated knowledge entity to be loaded into the health-preserving exercise knowledge section (the first target associated knowledge entity).

[0063] Based on the logical relationship between the associated knowledge entity to be expanded and the knowledge concept entity (special "qi"), the server determines the knowledge concept entity to be expanded. In this example, since the specific steps of the ancient health-preserving exercises are closely related to the operation and regulation of the special "qi" in the human body, the knowledge concept entity to be expanded may be a concept entity related to the special operation trajectory and regulation mechanism of the human meridian qi and blood in the health-preserving exercises.

[0064] Then, the server performs an association analysis on the knowledge concept entity to be expanded. The server will check the structure of the entire TCM knowledge graph and analyze the relationship between the knowledge concept entity to be expanded and other knowledge entities. For example, analyze its relationship with other organs, emotions, etc. of the human body. During the analysis process, the server will mine potential associated elements related to the knowledge concept entity to be expanded, and generate target-associated knowledge entities to be expanded after screening. For example, the association between certain movements in the new health-preserving exercise steps and specific meridian acupoints, as well as the association between these movements and emotional changes after regulating qi and blood, are identified as target-associated knowledge entities to be expanded.

[0065] Finally, the server loads these target-related knowledge entities to be expanded into the first target-related knowledge entity (health-preserving exercises knowledge section) of the TCM knowledge graph. If the first target-related knowledge entity is a category-related knowledge entity, like the health-preserving exercises knowledge section here, it was originally just a simple classification of related knowledge entities. After loading these target-related knowledge entities to be expanded, because its internal hierarchical structure becomes more complex, the category of this first target-related knowledge entity is adjusted to a group-related knowledge entity category. This completes the adjustment of this part of the architecture of the TCM knowledge graph, making the TCM knowledge graph more rich and accurately reflect TCM knowledge.

[0066] Taking the knowledge graph part related to herbal medicine processing methods as an example, the first target associated knowledge entity may be the knowledge block of temperature control in the processing process of a certain type of herbal medicine (category associated knowledge entity), and the associated knowledge entity to be expanded is the related knowledge entity that requires special adjustment of the processing temperature of a newly discovered herbal medicine under a special geographical environment. Based on the logical relationship between the associated knowledge entity to be expanded and the knowledge concept entity (herbal medicine processing temperature control), the knowledge concept entity to be expanded is determined to be the concept entity of the relationship between the influence of special geographical environmental factors on the properties of herbal medicine and the processing temperature. This knowledge concept entity to be expanded is subjected to association analysis to generate target associated knowledge entities such as the relationship between the change of herbal medicine ingredients and temperature adjustment under special geographical environments, and then these target associated knowledge entities to be expanded are loaded into the temperature control knowledge block. If it meets the situation of hierarchical associated knowledge entities, the category of the temperature control knowledge block becomes the category of group associated knowledge entities, thereby completing the adjustment of the TCM knowledge graph architecture in this part and generating the target TCM knowledge graph.

[0067] In a possible implementation, when the architecture adjustment instruction includes the associated knowledge entity destruction information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: A second target associated knowledge entity is obtained from the associated knowledge entity destruction information. The second target associated knowledge entity is an associated knowledge entity to be destroyed in the traditional Chinese medicine knowledge graph.

[0068] From the TCM knowledge graph, the second target-related knowledge entity is destroyed, the structural adjustment of the TCM knowledge graph is completed, and the target TCM knowledge graph is generated.

[0069] In a possible implementation, when the architecture adjustment instruction includes the associated knowledge entity migration information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: A third target associated knowledge entity and a fourth target associated knowledge entity are obtained from the associated knowledge entity migration information. The third target associated knowledge entity and the fourth target associated knowledge entity are associated knowledge entities that do not have a derivative relationship in the TCM knowledge graph, and the third target associated knowledge entity and the fourth target associated knowledge entity are determined by the knowledge section corresponding to the TCM knowledge graph.

[0070] In the TCM knowledge graph, the third target-related knowledge entity and the fourth target-related knowledge entity are migrated to complete the structural adjustment of the TCM knowledge graph and generate the target TCM knowledge graph.

[0071] In this embodiment, when the architecture adjustment instruction includes the destruction information of the associated knowledge entity, the server will perform specific operations according to the instruction. Taking the special "qi" related part in the TCM knowledge graph as an example, with the in-depth development of TCM research, it may be found that some associated knowledge entities in the previous graph are wrong or do not conform to the current TCM theoretical system. At this time, the server obtains the second target associated knowledge entity from the associated knowledge entity destruction information, and this second target associated knowledge entity is the associated knowledge entity to be destroyed in the TCM knowledge graph. For example, in the special "qi" related knowledge, there was an associated knowledge entity that claimed that a special stone worn on the body can directly regulate this special "qi" of the human body, but after rigorous TCM theoretical research and practical verification, it was found that this statement lacked basis. Then this associated knowledge entity about the special stone and the special "qi" is determined as the second target associated knowledge entity. Then, the server directly destroys this second target associated knowledge entity from the TCM knowledge graph. During the destruction process, the server will adjust the relevant connection relationship to ensure the coherence and logic of the graph. After completing this operation, the architecture adjustment of the TCM knowledge graph is ended, thereby generating the target TCM knowledge graph.

[0072] Let's look at the situation where the architecture adjustment instruction includes the migration information of associated knowledge entities. Still taking the special "qi" related knowledge graph as an example, assuming that there are associated knowledge entities about the special "qi" in the traditional folk health concept and the associated knowledge entity in the orthodox Chinese medicine classic theory in the graph. These two associated knowledge entities are the third target associated knowledge entity and the fourth target associated knowledge entity. They are associated knowledge entities that do not have a derivative relationship in the Chinese medicine knowledge graph, and are determined by the knowledge section corresponding to the Chinese medicine knowledge graph. In the traditional folk health concept, the special "qi" may have a special connection with some specific folk activities, while in the orthodox Chinese medicine classic theory, the special "qi" is more related to the theory of the human body's internal organs and meridians. With the demand for the integration of the Chinese medicine knowledge system, the server obtains these two target associated knowledge entities from the associated knowledge entity migration information and performs the migration operation in the Chinese medicine knowledge graph. For example, the special "qi" related knowledge entities in the traditional folk health concept are migrated to the knowledge section that is more suitable for the combination of folk culture and traditional Chinese medicine, and the special "qi" related knowledge entities in the orthodox traditional Chinese medicine classic theory are migrated to the viscera and meridian related knowledge section that can better reflect the core of traditional Chinese medicine theory, so that the classification of knowledge is more reasonable. During the migration process, the server will rebuild their relationship with the surrounding knowledge entities to ensure the integrity of the graph structure. After the migration operation is completed, the structural adjustment of the traditional Chinese medicine knowledge graph is completed, and the target traditional Chinese medicine knowledge graph is generated.

[0073] Similarly, similar operations are followed for the knowledge graph related to herbal preparation methods. If there is an associated knowledge entity that claims that a certain herb has magical effects that are inconsistent with TCM theory after preparation, this associated knowledge entity is determined to be the second target associated knowledge entity that needs to be destroyed, and the server will destroy it from the graph. If there are associated knowledge entities of herbal preparation methods in different regional theories that need to be adjusted, these associated knowledge entities are determined to be the third target associated knowledge entities and the fourth target associated knowledge entities, and the server will migrate them to the appropriate knowledge sections, thereby completing the architecture adjustment and generating the target TCM knowledge graph.

[0074] In a possible implementation, after step S120, the method further includes: Step B110, obtaining source associated knowledge entities from the TCM knowledge graph. The source associated knowledge entities are associated knowledge entities in the TCM knowledge graph that are used to trigger graphical representation.

[0075] Step B120: determining adjustment restriction information based on the source-related knowledge entity.

[0076] Step B130, loading the adjustment restriction information into each associated knowledge entity in the TCM knowledge graph, thereby ending the restriction on the TCM knowledge graph and generating a restricted TCM knowledge graph.

[0077] Wherein, step A120 includes: When the adjustment restriction information is in an open state, the restricted TCM knowledge graph is adjusted based on the adjustment instruction, and the adjustment restriction information is configured to be in an adjusted state.

[0078] When the adjustment of the restricted TCM knowledge graph is completed, the target TCM knowledge graph is generated, and the adjustment restriction information is configured to be in the open state.

[0079] In this embodiment, after completing the association analysis of each knowledge concept entity in the traditional Chinese medicine knowledge architecture and obtaining the associated knowledge entity, semantic relationship information and knowledge feature vector information corresponding to each knowledge concept entity, the server starts a new operation process.

[0080] The server obtains source-related knowledge entities from the TCM knowledge graph. Taking the special TCM knowledge graph related to "qi" as an example, it is assumed that this source-related knowledge entity is "qi nixing". This "qi nixing" has a unique position in the entire TCM knowledge graph. It has a complex relationship with many other knowledge entities, and due to its key role in TCM theory (such as its close relationship with human diseases, meridian qi and blood circulation, etc.), it is determined as an associated knowledge entity for triggering graph representation.

[0081] Based on this source-related knowledge entity "Qi Retrograde", the server determines the adjustment restriction information. For example, because "Qi Retrograde" mainly involves the knowledge related to the circulation of Qi and blood and meridians in the human body, the server will determine the relevant adjustment restriction information based on its related classical theories of traditional Chinese medicine, existing research results, and actual clinical experience. This adjustment restriction information may stipulate that when adjusting knowledge entities directly related to "Qi Retrograde" (such as knowledge entities such as symptoms and conditioning methods related to Qi Retrograde), it is necessary to follow specific principles of TCM Qi and blood theory, and its core semantic relationship information and knowledge feature vector information cannot be changed at will.

[0082] Then, the server loads this adjustment restriction information into each associated knowledge entity in the TCM knowledge graph. For example, for the disease knowledge entity "heart pulse is not smooth" related to "qi retrograde", after loading the adjustment restriction information, in the subsequent graph adjustment process, if you want to adjust "heart pulse is not smooth", you need to refer to this adjustment restriction information. In this way, the restriction of the TCM knowledge graph is completed, and the restricted TCM knowledge graph is generated.

[0083] When there is an adjustment instruction to adjust the TCM knowledge graph to generate a target TCM knowledge graph, if the adjustment restriction information is in an open state, the server can adjust the restricted TCM knowledge graph based on the adjustment instruction, and configure the adjustment restriction information to the adjustment state. For example, suppose the adjustment instruction is to update a conditioning method knowledge entity related to "Qi retrograde". Since the adjustment restriction information is in an open state, the server can adjust this conditioning method knowledge entity based on new research results or clinical experience. During the adjustment process, the adjustment restriction information is in an adjustment state, which means that the server must follow the basic principles of the previously determined restriction information when operating, such as not violating the TCM Qi and blood theory.

[0084] When the adjustment of the restricted TCM knowledge graph is completed, the server generates the target TCM knowledge graph and reconfigures the adjustment restriction information to an open state. For example, after completing the update of the knowledge entity of the conditioning method related to "Qi Nixing", the entire graph is adjusted. At this time, the generated target TCM knowledge graph reflects the latest TCM knowledge state, and the adjustment restriction information is restored to an open state for possible further adjustment operations in the future.

[0085] The same situation also applies to the part of the knowledge graph related to herbal preparation methods. If the source associated knowledge entity is a key step in the preparation process of a certain herb, the adjustment restriction information is determined based on this key step (such as following a specific herbal property conversion theory, etc.), and it is loaded into each associated knowledge entity to form a restricted knowledge graph. When there is an adjustment instruction and the adjustment restriction information is in an open state, the restricted graph is adjusted (such as updating the relationship with other knowledge entities in the preparation process, etc.). After the adjustment is completed, the target knowledge graph is generated and the adjustment restriction information is restored to an open state.

[0086] In a possible implementation, the knowledge concept entity in the TCM knowledge architecture includes basic information of traditional Chinese medicine.

[0087] Step S130 includes: Step S131, walking the TCM knowledge graph, and determining associated knowledge concept entities based on the semantic relationship information in each associated knowledge entity.

[0088] Step S132, obtaining the basic information of traditional Chinese medicine corresponding to the associated knowledge concept entity.

[0089] Step S133, based on the knowledge feature vector information and the traditional Chinese medicine basic information corresponding to the associated knowledge concept entity, the feature vector corresponding to the current display node in each associated knowledge entity is represented graphically to generate a current Chinese medicine knowledge display node. The current Chinese medicine knowledge display node belongs to the visual display data flow of the Chinese medicine knowledge graph.

[0090] Step S134, continue to walk the TCM knowledge graph, and graphically represent the feature vector corresponding to the next display node until the graphical representation of the feature vectors corresponding to each display node is completed, and the graphical representation is completed to generate a visual display data stream of the TCM knowledge graph.

[0091] In this embodiment, since the knowledge concept entities in the TCM knowledge architecture contain basic information of traditional Chinese medicine, the server begins to roam the TCM knowledge graph for graphical representation to generate a visual display data stream.

[0092] Taking the special "qi" related knowledge as an example, the server starts to walk through the TCM knowledge graph and determines the associated knowledge concept entity based on the semantic relationship information in each associated knowledge entity. For example, for the associated knowledge entity "qi retrograde", based on its semantic relationship information (such as qi retrograde will affect the circulation of qi and blood in the heart), it can be determined that the associated knowledge concept entity is "qi". Because in the TCM theoretical system, "qi" is a basic concept, and "qi retrograde" is a special state of "qi", and there is a close semantic connection between the two.

[0093] Then, the server obtains the basic information of traditional Chinese medicine corresponding to the associated knowledge concept entity. For the associated knowledge concept entity "qi", the basic information of traditional Chinese medicine includes the source of qi (such as innate qi comes from parents, acquired qi comes from diet, breathing, etc.), the classification of qi (such as primordial qi, ancestral qi, nutrient qi, defensive qi, etc.) and the basic operation rules of qi in the human body (such as qi running along the meridians and interacting with the internal organs, etc.).

[0094] Based on the knowledge feature vector information and the traditional Chinese medicine basic information corresponding to the associated knowledge concept entity, the feature vector corresponding to the current display node in each associated knowledge entity is represented graphically to generate the current Chinese medicine knowledge display node. For the associated knowledge entity "Qi retrograde", its knowledge feature vector information may include the triggering factors of Qi retrograde (such as excessive emotions, invasion of external evil, etc.), the impact of Qi retrograde on the human body (such as poor heart pulse, dysfunction of internal organs, etc.). Combined with the traditional Chinese medicine basic information of "Qi", it is represented in the graph. For example, in the visual display data flow, the "Qi retrograde" node can be represented by a specific icon (such as a red arrow indicating the retrograde direction) and connected to the "Qi" node by a line. The line is marked with the interference of Qi retrograde on the normal operation of Qi, and the text information such as its triggering factors and impacts are displayed around the "Qi retrograde" node. In this way, the current Chinese medicine knowledge display node is generated, which belongs to the visual display data flow of the Chinese medicine knowledge graph.

[0095] Next, the server continues to walk through the TCM knowledge graph and graphically represents the feature vector corresponding to the next display node. Assuming that the next associated knowledge entity is "heart pulse is not smooth", the associated knowledge concept entity is determined to be "heart pulse" based on its semantic relationship information (qi retrograde leads to heart pulse is not smooth). Obtain the basic traditional Chinese medicine information corresponding to "heart pulse", including the direction of the meridians of the heart pulse (starting from the heart, belonging to the heart system, etc.), the intersection of the heart pulse with other meridians, and the characteristics of the heart pulse qi and blood circulation. Based on the knowledge feature vector information of "heart pulse is not smooth" (such as symptoms such as chest tightness, palpitations, specific conditions of qi and blood stasis, etc.) and the basic traditional Chinese medicine information of "heart pulse", the node "heart pulse is not smooth" is represented in the graph. For example, a lighter color line is used to represent the state of poor heart pulse qi and blood, and specific symptoms are listed at the node to generate a new TCM knowledge display node.

[0096] In this way, the server continues to walk the TCM knowledge graph and graphically represents the feature vectors corresponding to each display node until the graphical representation of the feature vectors corresponding to each display node is completed, and the graphical representation is completed to generate a complete visual display data stream of the TCM knowledge graph. This data stream is like a dynamic TCM knowledge map, which fully displays the complex relationships, features, and other contents between various knowledge entities in the TCM knowledge system.

[0097] The same is true for the knowledge related to herbal preparation methods. For example, for the associated knowledge entity "herbal medicine property transformation", the associated knowledge concept entity is determined to be the herb itself based on the semantic relationship information, and the traditional Chinese medicine basic information of the herb (such as the nature, taste, meridians, efficacy and indications of the herb) is obtained. Combined with the knowledge feature vector information of "herbal medicine property transformation" (such as the conditions for transformation, changes in medicinal properties after transformation, etc.), the "herbal medicine property transformation" node is represented in a graph to generate a display node. Then continue to walk the graph and perform similar operations on the next display node related to herbal preparation until the generation of the visual display data flow of the entire herbal medicine preparation related knowledge in the Chinese medicine knowledge graph is completed.

[0098] For example, in a possible implementation, step S120 includes: Step S121, analyzing the overall structure of the TCM knowledge framework and the contextual relationship of each knowledge concept entity in the overall structure, and determining the initial semantic domain scope involved by each knowledge concept entity.

[0099] In step S122, based on the determined initial semantic domain, the data of classical Chinese medicine literature is searched and screened to mine a set of text descriptions related to the initial semantic domain.

[0100] Step S123, parse the text description set, analyze the basic logical relationship between the grammatical structure components and the TCM connotation of each word and phrase in the sentence, identify the potential associated elements related to the knowledge concept entity based on the basic logical relationship between the grammatical structure components and the TCM connotation of each word and phrase, and generate a list of potential associated elements.

[0101] Step S124, classify the list of potential associated elements according to the logic and classification standards of the traditional Chinese medicine knowledge system, evaluate the degree of association between the elements and the knowledge concept entities for each classification category, screen out the elements that are closely associated and conform to the logic of traditional Chinese medicine knowledge, and generate a subset of valid associated elements, wherein the evaluation of the degree of association is based on the causal relationship and functional connection in traditional Chinese medicine theory.

[0102] Step S125, each valid associated element in the valid associated element subset is determined as an associated knowledge entity, and based on the basic logical relationship between the grammatical structure component elements in the text description set, the semantic relationship information between each associated knowledge entity and the knowledge concept entity is constructed, and one or more associated knowledge entities corresponding to the knowledge concept entity and the semantic relationship information between the associated knowledge entity and the knowledge concept entity are generated.

[0103] Step S126: for each associated knowledge entity, extract the knowledge feature vector information of the associated knowledge entity according to the attribute feature definition in the traditional Chinese medicine knowledge system.

[0104] In this embodiment, continuing to take the special knowledge concept entity "qi" as an example, the server first analyzes the overall structure of the TCM knowledge architecture and the contextual relationship of "qi" in this overall structure. In the TCM knowledge architecture, the overall structure includes multiple sections such as human physiological structure (viscera, meridians, etc.), pathological phenomena, health care, and drug treatment. The concept entity "qi" is in a core position and is closely related to the viscera in the human physiological structure. For example, "qi" is the basic substance to maintain the functional activities of the viscera; it is closely related to the meridians and is an important material basis for the operation of the meridians; in terms of pathological phenomena, the disorder of "qi" (such as qi deficiency, qi reversal, etc.) is often an important factor in causing diseases; in health care, many principles of various health-preserving exercises (such as Tai Chi, etc.) are based on the regulation of "qi"; in drug treatment, the efficacy of drugs is often related to the regulation of "qi" (such as qi-tonifying drugs, qi-moving drugs, etc.). Through such overall structural analysis and the combing of the contextual relationship of "qi", the initial semantic domain scope involved in the knowledge concept entity "qi" is determined. This scope covers the generation of Qi (both innate and acquired), the movement patterns of Qi (the movement paths and directions between meridians and internal organs, etc.), the functions of Qi (such as promotion, warmth, defense, etc.), the abnormal manifestations of Qi (such as Qi deficiency, Qi disorder, etc.) and the relationship with other physiological and pathological factors.

[0105] Next, based on the determined initial semantic domain of "qi", the server searches and screens the data of classical Chinese medicine literature. These classical Chinese medicine literature data come from a wide range of sources, including various ancient Chinese medicine classics, local Chinese medicine inheritance literature, etc. During the search process, the server will search for literature fragments containing content related to the generation, operation, function, abnormal performance, and mutual relationship of "qi". For example, from the Yellow Emperor's Internal Classic, the sentence "People receive qi from food, food enters the stomach, and is transmitted to the lungs, and the five internal organs and six bowels all receive qi" is selected to describe the relationship between the generation of qi and the internal organs; from the Difficult Classic, the relevant description of the path of qi operation "the movement of qi is like the flow of water, and it cannot stop" is found. Through such search and screening, a collection of text descriptions related to the initial semantic domain of "qi" is excavated.

[0106] Then, the server parses this text description set. Taking the sentence "People receive Qi from food, food enters the stomach, and is transmitted to the lungs. The five internal organs and six bowels all receive Qi" as an example, the server analyzes the basic logical relationship between the grammatical structure elements in the sentence. In this sentence, "People receive Qi from food" is a statement relationship, indicating that the source of Qi for people is food; "food enters the stomach" is an action relationship, describing the process of food entering the stomach; "to be transmitted to the lungs" indicates a transmission relationship, that is, the Qi transformed from food is transmitted from the stomach to the lungs; "the five internal organs and six bowels all receive Qi" is a relationship between the whole and the part, indicating that the five internal organs and six bowels all receive Qi from this transmission process. At the same time, the server analyzes the TCM connotation of each word and phrase. In the connotation of traditional Chinese medicine, "gu" is not just ordinary food. It has five flavors (sour, bitter, sweet, pungent, and salty). Different flavors have specific affinity with the five internal organs, such as sweet enters the spleen. This relationship between the five flavors and the five internal organs will affect the generation and transformation of qi; "stomach" is an important internal organ in traditional Chinese medicine, the sea of ​​water and grain, and an important place for the initial digestion of food and its transformation into qi; "lungs" play a key role in the transmission and distribution of qi in addition to their respiratory function. Based on the basic logical relationship between the elements of this grammatical structure and the connotation of each word and phrase in traditional Chinese medicine, the server identifies potential related elements related to the knowledge concept entity "qi". Here, "gu", "stomach", "lungs", "five internal organs and six bowels", etc. are all potential related elements, and these elements are generated into a list of potential related elements.

[0107] Afterwards, the server classifies this list of potential related elements according to the logic and classification standards of the TCM knowledge system. For example, "gu" belongs to the diet and nutrition category, "stomach" and "lung" belong to the viscera category, and "five internal organs and six bowels" belong to the viscera system category. For each classification category, evaluate the degree of association between the elements and the knowledge concept entity "qi". During the evaluation process, the evaluation is based on the causal relationship and functional connection in TCM theory. Taking "stomach" as an example, from the perspective of causality, the normal digestive function of the stomach directly affects the generation of qi. The stomach can normally decompose water and grain, and the qi can be sufficient, so the degree of association is very high; from the perspective of functional connection, the stomach is one of the sources of qi generation and has a clear functional connection with qi. For "gu", it is the material basis for the generation of qi. Without grain, there is no source of qi, and it has a strong causal relationship with qi. As for the overall concept of "five internal organs and six bowels", from the perspective of functional connection, they are the recipients and users of qi. Qi plays a role in promoting and warming the five internal organs and six bowels, and the degree of association is also very high. After such an evaluation, elements that are closely related and consistent with the logic of traditional Chinese medicine knowledge, such as "grain", "stomach", "lungs", "five internal organs", etc., are screened out to generate a subset of valid related elements.

[0108] Each valid associated element in the valid associated element subset is then determined as an associated knowledge entity. For example, "Valley", "Stomach", "Lung", and "Five Internal Organs and Six Bowels" have all become associated knowledge entities associated with the knowledge concept entity "Qi". Then, based on the basic logical relationship between the grammatical structure component elements in the text description set, the semantic relationship information between each associated knowledge entity and the knowledge concept entity is constructed. For "Valley" and "Qi", the semantic relationship information is that valley is the source of Qi; for "Stomach" and "Qi", the semantic relationship information is that stomach is an important place for Qi generation; for "Lung" and "Qi", the semantic relationship information is that lung is an important organ for Qi transmission and distribution; for "Five Internal Organs and Six Bowels" and "Qi", the semantic relationship information is that five internal organs and six bowels are the places where Qi functions. In this way, multiple associated knowledge entities corresponding to the knowledge concept entity "Qi" and their semantic relationship information with "Qi" are generated.

[0109] Finally, for each associated knowledge entity, the knowledge feature vector information of the associated knowledge entity is extracted according to the attribute feature definition in the TCM knowledge system. For the associated knowledge entity "gu", its knowledge feature vector information includes the types of gu (such as polished rice, wheat, etc.), the five flavors of gu (different gu have different flavors, such as polished rice tastes sweet, etc.), the effects of gu (such as some gu have the effects of strengthening the spleen and replenishing qi, etc.). For the associated knowledge entity "stomach", its knowledge feature vector information covers the physiological structure characteristics of the stomach (such as the upper opening of the stomach is the cardia, the lower opening is the pylorus, etc.), the functional characteristics of the stomach (such as the main receiving and decomposing of water and grain, etc.), the pathological manifestations of the stomach (such as epigastric pain, vomiting, etc.). For the associated knowledge entity "lung", the knowledge feature vector information includes the anatomical structure of the lung (such as the lung is located in the chest cavity, etc.), the physiological functions of the lung (mainly responsible for breathing, mainly for spreading and descending, etc.), and the pathological changes of the lung (such as coughing, asthma, etc.). For the associated knowledge entity of "five internal organs and six bowels", the knowledge feature vector information includes the names, functions, relationships (such as the mutual promotion and restraint relationship between the five internal organs), and their specific manifestations in the circulation of qi and the performance of functions.

[0110] Similarly, take the knowledge concept entity of "herbal medicine processing method" as an example. The server analyzes the overall structure of the TCM knowledge architecture and finds that the herbal medicine processing method is related to the nature and efficacy of the herbal medicine itself, as well as the human body's acceptance of the processed herbal medicine. The initial semantic domain scope is determined to include the types of processing methods, the changes in the nature of the herbal medicine caused by processing, and the connection between processing and TCM theory (such as five flavors, meridians, etc.). Then, based on this scope, the classic TCM literature is retrieved and screened to mine relevant text description sets, such as "When using croton, peel the carpel membrane and take the frost" and other sentences describing the processing process. These sentences are parsed, the grammatical structure and lexical TCM connotations are analyzed, and potential related elements (such as croton, carpel membrane, frost, etc.) are identified to generate a list. These elements are classified according to the TCM knowledge logic, the degree of correlation with "herbal medicine processing method" is evaluated, and a subset of valid related elements (such as croton, frost, etc.) is screened. The subset elements are identified as associated knowledge entities, and semantic relationship information is constructed (such as the relationship between croton and processing methods is that croton is the processed object, frost is the product after processing, etc.), and finally the knowledge feature vector information of each associated knowledge entity is extracted (such as the toxicity and meridians of croton, the medicinal properties of frost, etc.).

[0111] Figure 2 The hardware structure of the TCM knowledge graph construction system 100 based on AI semantic understanding for implementing the TCM knowledge graph construction method based on AI semantic understanding provided by an embodiment of the present invention is shown as follows: Figure 2 As shown, the TCM knowledge graph construction system 100 based on AI semantic understanding may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0112] The machine-readable storage medium 120 can store data and / or instructions. In some embodiments, the machine-readable storage medium 120 can store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 can store data and / or instructions that the TCM knowledge graph construction system 100 based on AI semantic understanding uses to execute or use to complete the exemplary method described in the present invention.

[0113] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the method for constructing a traditional Chinese medicine knowledge graph based on AI semantic understanding as described in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected via the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0114] The specific implementation process of the processor 110 can be found in the various method embodiments executed by the above-mentioned TCM knowledge graph construction system 100 based on AI semantic understanding. The implementation principles and technical effects are similar and will not be repeated in this embodiment.

[0115] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer executable instructions are preset. When the processor executes the computer executable instructions, the above-mentioned method for constructing a traditional Chinese medicine knowledge graph based on AI semantic understanding is implemented.

[0116] It should be noted that in order to simplify the description of the present invention and thus facilitate the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A method for constructing a TCM knowledge graph based on AI semantic understanding, characterized in that: The method comprises: Acquire TCM classic literature data, and construct a TCM knowledge framework from the TCM classic literature data; the TCM classic literature data is a data source corresponding to TCM theory, the TCM knowledge framework is a feature vector of a TCM knowledge graph to be constructed, and the TCM knowledge framework includes one or more knowledge concept entities; Performing association analysis on each knowledge concept entity in the TCM knowledge architecture, generating one or more associated knowledge entities corresponding to each knowledge concept entity in the TCM knowledge graph, and semantic relationship information and knowledge feature vector information corresponding to each associated knowledge entity in the one or more associated knowledge entities, wherein the semantic relationship information is the logical relationship between each associated knowledge entity and the knowledge concept entity, and the knowledge feature vector information is the specific TCM knowledge attribute feature represented by each associated knowledge entity; Wandering the TCM knowledge graph, based on the semantic relationship information and the knowledge feature vector information, graphically represent each associated knowledge entity, and generate a visual display data stream of the TCM knowledge graph.

2. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 1, characterized in that: Before walking the TCM knowledge graph and graphically representing each associated knowledge entity based on the semantic relationship information and the knowledge feature vector information to generate a visual display data stream of the TCM knowledge graph, the method further includes: Obtaining an adjustment instruction for the TCM knowledge graph; the adjustment instruction is instruction information for adjusting the TCM knowledge graph; Adjust the TCM knowledge graph based on the adjustment instruction to generate a target TCM knowledge graph; Wherein, the walking of the TCM knowledge graph, based on the semantic relationship information and the knowledge feature vector information, performing a graph representation of each associated knowledge entity, and generating a visual display data stream of the TCM knowledge graph, includes: Wander the target TCM knowledge graph, and based on the target semantic relationship information and the target knowledge feature vector information, graphically represent each target-related knowledge entity in the target TCM knowledge graph, and generate a visual display data stream of the TCM knowledge graph; the target semantic relationship information and the target knowledge feature vector information correspond to each target-related knowledge entity.

3. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 2, characterized in that: The adjustment instruction is a category knowledge adjustment instruction; the category knowledge adjustment instruction is instruction information for adjusting the attribute characteristics of TCM knowledge corresponding to the associated knowledge entity in the TCM knowledge graph; The step of adjusting the TCM knowledge graph based on the adjustment instruction to generate a target TCM knowledge graph includes: From the category knowledge adjustment instruction, obtain target associated knowledge entity and knowledge feature vector information to be adjusted; the target associated knowledge entity is an associated knowledge entity in the traditional Chinese medicine knowledge graph whose knowledge feature is to be adjusted; Obtaining target knowledge feature vector information corresponding to the target associated knowledge entity; Based on the knowledge feature vector information to be adjusted, the target knowledge feature vector information is updated, thereby completing the adjustment of the TCM knowledge graph and generating the target TCM knowledge graph.

4. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 2, characterized in that: The adjustment instruction is an architecture adjustment instruction; the architecture adjustment instruction is instruction information for adjusting the architecture of the TCM knowledge graph; The step of adjusting the TCM knowledge graph based on the adjustment instruction to generate a target TCM knowledge graph includes: Based on the architecture adjustment instruction, the architecture of the TCM knowledge graph is adjusted to generate the target TCM knowledge graph; the architecture adjustment instruction includes at least one of associated knowledge entity extension information, associated knowledge entity destruction information and associated knowledge entity migration information; Among them, the associated knowledge entity extension information is the feature vector for expanding the associated knowledge entity into the traditional Chinese medicine knowledge graph, the associated knowledge entity destruction information is the feature vector for destroying the associated knowledge entity from the traditional Chinese medicine knowledge graph, and the associated knowledge entity migration information is the feature vector for migrating the associated knowledge entity in the traditional Chinese medicine knowledge graph.

5. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 4, characterized in that: When the architecture adjustment instruction includes the associated knowledge entity extension information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: From the associated knowledge entity extension information, a first target associated knowledge entity and an associated knowledge entity to be extended are obtained, wherein the first target associated knowledge entity is a category associated knowledge entity or a group associated knowledge entity in the TCM knowledge graph for which feature vector extension is to be performed; and the associated knowledge entity to be extended is an initial associated knowledge entity to be loaded into the first target associated knowledge entity; Determine the knowledge concept entity to be expanded based on the logical relationship between the associated knowledge entity to be expanded and the knowledge concept entity; Performing association analysis on the knowledge concept entity to be expanded to generate a target associated knowledge entity to be expanded; The target associated knowledge entity to be expanded is loaded into the first target associated knowledge entity of the TCM knowledge graph, the structural adjustment of the TCM knowledge graph is completed, and the target TCM knowledge graph is generated, the first target associated knowledge entity and the target associated knowledge entity to be expanded are the same-level associated knowledge entities or the hierarchical associated knowledge entities, and when the first target associated knowledge entity is a category associated knowledge entity, and the first target associated knowledge entity and the target associated knowledge entity to be expanded are hierarchical associated knowledge entities, the category of the first target associated knowledge entity after adjustment is a group associated knowledge entity category.

6. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 4, characterized in that: When the architecture adjustment instruction includes the associated knowledge entity destruction information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: Acquire a second target associated knowledge entity from the associated knowledge entity destruction information; the second target associated knowledge entity is an associated knowledge entity to be destroyed in the traditional Chinese medicine knowledge graph; From the TCM knowledge graph, the second target-related knowledge entity is destroyed, the structural adjustment of the TCM knowledge graph is completed, and the target TCM knowledge graph is generated.

7. The method for constructing a TCM knowledge graph based on AI semantic understanding according to claim 4, characterized in that: When the architecture adjustment instruction includes the associated knowledge entity migration information, adjusting the architecture of the TCM knowledge graph based on the architecture adjustment instruction to generate the target TCM knowledge graph includes: From the associated knowledge entity migration information, a third target associated knowledge entity and a fourth target associated knowledge entity are obtained; the third target associated knowledge entity and the fourth target associated knowledge entity are associated knowledge entities that do not have a derivative relationship in the traditional Chinese medicine knowledge graph, and the third target associated knowledge entity and the fourth target associated knowledge entity are determined by the knowledge section corresponding to the traditional Chinese medicine knowledge graph; In the TCM knowledge graph, the third target-related knowledge entity and the fourth target-related knowledge entity are migrated to complete the structural adjustment of the TCM knowledge graph and generate the target TCM knowledge graph.

8. The method for constructing a TCM knowledge graph based on AI semantic understanding according to any one of claims 2 to 7, characterized in that: After performing association analysis on each knowledge concept entity in the TCM knowledge architecture to generate one or more associated knowledge entities corresponding to each knowledge concept entity in the TCM knowledge graph, as well as semantic relationship information and knowledge feature vector information corresponding to each associated knowledge entity in the one or more associated knowledge entities, the method further includes: From the TCM knowledge graph, a source associated knowledge entity is obtained; the source associated knowledge entity is an associated knowledge entity in the TCM knowledge graph used to trigger a graphical representation; Determining adjustment restriction information based on the source-related knowledge entity; Loading the adjustment restriction information into each associated knowledge entity in the TCM knowledge graph, thereby ending the restriction on the TCM knowledge graph and generating a restricted TCM knowledge graph; The step of adjusting the TCM knowledge graph based on the adjustment instruction to generate a target TCM knowledge graph includes: When the adjustment restriction information is in an open state, adjusting the restricted TCM knowledge graph based on the adjustment instruction, and configuring the adjustment restriction information to be in an adjustment state; When the adjustment of the restricted TCM knowledge graph is completed, the target TCM knowledge graph is generated, and the adjustment restriction information is configured to be in the open state.

9. The method for constructing a TCM knowledge graph based on AI semantic understanding according to any one of claims 2 to 7, characterized in that: The knowledge concept entities in the TCM knowledge framework include basic information of traditional Chinese medicine; The walking the TCM knowledge graph, based on the semantic relationship information and the knowledge feature vector information, performing a graph representation on each associated knowledge entity, and generating a visual display data stream of the TCM knowledge graph, includes: Walking the TCM knowledge graph, and determining associated knowledge concept entities based on the semantic relationship information in each associated knowledge entity; Acquire the basic information of traditional Chinese medicine corresponding to the associated knowledge concept entity; Based on the knowledge feature vector information and the traditional Chinese medicine basic information corresponding to the associated knowledge concept entity, the feature vector corresponding to the current display node in each associated knowledge entity is represented graphically to generate a current Chinese medicine knowledge display node; the current Chinese medicine knowledge display node belongs to the visual display data flow of the Chinese medicine knowledge graph; Continue to roam the TCM knowledge graph, and graphically represent the feature vector corresponding to the next display node until the graphical representation of the feature vectors corresponding to each display node is completed, and the graphical representation is completed to generate a visual display data stream of the TCM knowledge graph.

10. A TCM knowledge graph construction system based on AI semantic understanding, characterized in that: The system for constructing a TCM knowledge graph based on AI semantic understanding includes a processor and a memory, wherein the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the method for constructing a TCM knowledge graph based on AI semantic understanding as described in any one of claims 1 to 9.

Images