Construction algorithm of disease-dish data knowledge graph

Abstract

The invention discloses a construction algorithm of a disease-dish data knowledge graph. The construction algorithm comprises the following steps: determining a disease-dish knowledge model in the knowledge graph; collecting structured, unstructured and semi-structured original dish data and disease knowledge, and storing the data and the knowledge; extracting entities, relationships and entity attributes from the data by using an information extraction technology; knowledge fusion: fusing the extracted information, including filtering redundant information and merging similar entities, so as to obtain unified knowledge representation; performing knowledge calculation, including quality evaluation, missing value processing and knowledge updating; knowledge generated in the process is stored in a graph database so as to be convenient to query and use.

Description

Technical Field

[0001] The present application relates to the field of knowledge graph and knowledge base construction, and in particular to a novel algorithm for constructing a disease-dish data knowledge graph. Background Art

[0002] The concept of knowledge graph was first proposed by Google in 2012 to enhance the search engine's capabilities, not only to provide links, but also to directly answer questions and explain facts. By expressing multiple relationships and attributes between entities through a graph structure, knowledge graphs are derived from the concepts of semantic web and linked data, and aim to create a knowledge base that can be understood and processed by machines. Since then, the development of knowledge graphs has promoted the advancement of information retrieval from simple queries to complex problem solving and decision support systems.

[0003] In the medical field, knowledge graphs have become an important tool to promote the advancement of health and medical information technology. It can integrate large and complex medical data from different sources, providing strong support for medical research, clinical decision support and patient care. For example, doctors can use knowledge graphs to quickly obtain the latest medical research results to help develop personalized treatment plans; researchers can discover potential disease associations by analyzing large-scale data sets, thereby accelerating the process of new drug development.

[0004] However, in the field of health care, there is relatively little work on organizing food and disease data into knowledge graphs. Although many studies have focused on how to use food nutrients to prevent or assist in the treatment of certain diseases, there is a lack of systematic methods to build a knowledge network that can reflect the deep connection between the two. This limits the ability to manage health based on eating habits and also hinders the development of customized dietary recommendations for specific populations.

[0005] In order to fill this gap, it is necessary to develop a new algorithm to build a dense disease-dish data knowledge graph. This graph should not only contain known medical knowledge, but also consider various factors involved in daily diet, such as ingredient selection, cooking methods, etc., and their long-term effects on human health. In addition, considering individual differences (such as age, gender, genetic background), the ideal solution should also be flexible enough to adapt to different user needs and provide personalized guidance.

[0006] In summary, there is an urgent need for a method that can effectively integrate food knowledge and disease knowledge to lay a solid data foundation for knowledge graph-based medical problem solving, user food preference analysis and recommendation, and other related applications. In this way, not only can the public's awareness of the importance of a reasonable diet be enhanced, but also more accurate and effective tools can be provided to medical service providers to help achieve the goal of "preventive treatment", that is, to reduce the occurrence and development of diseases through preventive measures. Summary of the invention

[0007] In response to the deficiencies of the prior art, this application proposes an algorithm for constructing a disease-dish data knowledge graph.

[0008] The technical solution adopted by the present application to solve the technical problem is to include the following steps: An algorithm for constructing a disease-dish data knowledge graph includes the following steps: Determine the disease-dish knowledge modeling in the knowledge graph; Collect and store structured, unstructured and semi-structured raw food data and disease knowledge; Extracting entities, relationships, and entity attributes from the data using information extraction techniques; Knowledge fusion, which integrates the extracted information, including filtering redundant information and merging similar entities to obtain a unified knowledge representation; Perform knowledge computation, including quality assessment, missing value processing, and knowledge updating; The knowledge generated by the above process is stored in a graph database for easy query and use.

[0009] In some embodiments, the knowledge modeling divides the knowledge graph into a pattern graph and a data graph, wherein the pattern graph is used to reflect the relationship between concepts, and the data graph is used to reflect the relationship between actual data.

[0010] In some embodiments, the information extraction technique includes using a large language model for entity recognition, relationship extraction, and attribute extraction.

[0011] In some embodiments, the large language model used is a model based on the Transformers architecture.

[0012] In some embodiments, the knowledge fusion further comprises: Entity linking: entity disambiguation and coreference resolution through similarity calculation; Knowledge merging: handling entity conflicts between data and schema layers from different sources and merging external knowledge bases with existing ones.

[0013] In some embodiments, it also includes creating a semantic ontology to construct entities, relationships, and attributes into a concept set or framework, so as to achieve classification and organization of data.

[0014] In some embodiments, the knowledge calculation step also involves using an inference engine to perform data reasoning to supplement the missing relationship values ​​between subjects in the knowledge graph and complete further knowledge discovery.

[0015] In some embodiments, the graph database is selected from Neo4J, Cayley, and ArangoDB to ensure efficient storage and query capabilities and enhance system response speed.

[0016] In some embodiments, a computer-readable storage medium is also included, on which a computer program is stored. When the program is executed by a processor, the method for constructing a disease-dish data knowledge graph according to any of the aforementioned embodiments is implemented.

[0017] The beneficial effects of this application are: through the above implementation methods, this application constructs an efficient, accurate, and continuously updated disease-dish data knowledge graph, providing strong data support for research and application in the field of medical health. Specific implementation effects include: 1. Efficient data query capability: By using graph databases for storage, users can quickly and intuitively query the complex relationship between diseases and dishes. The structure of graph databases makes the retrieval speed of related data much faster than traditional relational databases, greatly improving user experience and efficiency.

[0018] 2. Dynamic data update mechanism: It realizes the real-time update and expansion function of the knowledge graph. The present invention includes a continuous update mechanism that can regularly check the reliability and consistency of the data source to ensure that the knowledge graph is always up to date, thereby ensuring the cutting-edge and accuracy of the information provided.

[0019] 3. Improve data credibility: By performing quality assessment and knowledge reasoning on the extracted information, the information in the knowledge base is ensured to be accurate and reliable. This includes not only quality control of the original data, but also the supplementation of the missing parts of the relationship values ​​between the subjects in the knowledge graph, as well as further knowledge discovery process.

[0020] 4. Enhanced personalized services: The knowledge graph based on disease-dish data provides the possibility for personalized health management. By analyzing the user's eating habits, physical condition and personal preferences, customized dietary advice and services can be provided to help prevent or assist in the treatment of specific diseases and promote individual health.

[0021] The above specific implementation methods demonstrate the comprehensive implementation method of the present invention, which realizes an efficient, accurate and dynamically updated disease-dish knowledge graph through the integration of a large language model, a graph database and an ETL tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present application is further described below in conjunction with the accompanying drawings and embodiments.

[0023] Figure 1 This is a schematic diagram of the knowledge graph construction process of this application.

[0024] Figure 2 It is a schematic diagram of the dish-disease knowledge graph of this application.

[0025] Figure 3 This is a data graph sample of the dish-disease knowledge graph of this application.

[0026] Figure 4 It is a flowchart for resolving and merging conflicting and ambiguous entities in the product-disease knowledge graph of this application.

[0027] Figure 5 This is the effect diagram of the extraction of entities, attributes, and relationships between dishes and diseases in this application. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present application are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application. In addition, the technical solutions between the various embodiments can be combined with each other, but must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that the combination of such technical solutions does not exist and is not within the scope of protection of the present application.

[0029] Specific implementation methods of this application: 1. Step 1: Modeling disease-dish knowledge in the knowledge graph: Reference Figure 1 As shown, the disease-dish data knowledge graph construction process of this application follows 6 main steps: Knowledge modeling: First, define the relationships between concepts such as diseases, dishes, ingredients, nutrients, symptoms, and treatment plans, and model these concepts. This stage determines the structural framework of the knowledge graph, ensuring that all information in subsequent steps can be accurately mapped to the preset concept model.

[0030] Raw data acquisition and storage: Collect structured, unstructured, and semi-structured raw data from different sources and store them in appropriate databases. For example, structured data is stored in MySQL, unstructured data is stored in MongoDB, and semi-structured data is stored in NoSQL databases such as Cassandra.

[0031] Information extraction: Use a large language model (LLM) based on the Transformers architecture for entity recognition, relationship extraction, and attribute extraction. This step extracts valuable structured information from the collected data, laying the foundation for subsequent knowledge fusion.

[0032] Knowledge fusion: Logically attribute and filter out redundant entities, relationships, and attributes obtained during information extraction. Through entity linking and knowledge merging, possible conflicts or duplications are handled to ensure data consistency and integrity.

[0033] Knowledge computing: ontology construction, knowledge reasoning, quality assessment, etc. This includes creating semantic ontologies to organize data, using reasoning engines to supplement missing relationship values, and using specific indicators to evaluate the quality of knowledge.

[0034] Knowledge base storage: Ultimately, the processed knowledge is structured and stored in a graph database, such as Neo4J, Cayley, or ArangoDB, for efficient query and use.

[0035] 2. Step 2: Model the disease-dish knowledge in the knowledge graph: Pattern diagram and data diagram 2.1. Pattern diagram: It is a relatively abstract concept network, which mainly reflects the relationship between concepts. For example, the relationship between dishes, ingredients and nutrients, and the relationship between diseases, symptoms, treatment plans and examination methods.

[0036] Example: Donuts -> high sugar and fat content -> leads to high blood sugar and high blood lipids -> causes dizziness -> requires diet control.

[0037] Reference Figure 2 As shown in the figure, this diagram shows a relatively abstract concept network, which reflects the relationship between concepts, that is, a pattern diagram. It not only reflects the relationship between dishes, ingredients, and nutrients, but also covers the relationship between medical concepts such as diseases, symptoms, treatment plans, and examination methods.

[0038] For example: Donuts -> high sugar and fat content -> lead to high blood sugar and high blood lipids -> cause dizziness -> need to control diet Diabetic patients -> Recommend low-sugar diet -> Reduce the risk of complications This pattern diagram helps us understand the deep-seated internal laws between different concepts and provides a theoretical basis for the subsequent construction of specific data diagrams.

[0039] 2.2. Data graph: reflects the relationship between actual data. The deep internal rules are reflected by the concept graph. Define specific entities, relationships and attributes.

[0040] Dish data entity: dish (donut), nutrients (sugar, oil), ingredients (flour, butter).

[0041] Disease data entities: disease (hyperglycemia), symptoms (dizziness), treatment plan (diet control), examination method (blood sugar test), medicine (insulin), recommended diet (low-sugar diet).

[0042] Relationship: include (such as dishes include ingredients), cause (such as high-sugar diet causes high blood sugar), check (such as blood sugar test).

[0043] Tools: Protege for OWL, ontology editing tool.

[0044] Reference Figure 3 As shown in the figure, this diagram reflects the relationship between actual data and is a specific instantiation of the pattern diagram. It defines specific entities, relationships, and attributes, allowing us to see how diseases and dishes in the real world interact with each other. For example: Dish data entities: donuts (dish), sugar (nutrient element), flour (ingredients) Disease data entities: high blood sugar (disease), dizziness (symptom), diet control (treatment plan), blood sugar test (test method), insulin (drug), low sugar diet (recommended diet) In addition, the graph also shows various relationships, such as "includes" (e.g., dishes contain ingredients), "cause" (e.g., high-sugar diet causes high blood sugar), "check" (e.g., blood sugar test), etc. In this way, the complex relationship between diseases and dishes can be intuitively displayed, providing users with valuable health guidance.

[0045] 3. Step 3: Original data acquisition and storage: Structured data: stored in the relational database MySQL.

[0046] Unstructured data: stored in MongoDB.

[0047] Semi-structured data: stored in the corresponding NoSQL database.

[0048] Tools: MySQL database, MongoDB, Cassandra, etc.

[0049] 4. Step 4: Information extraction: Extraction of entities, relations and attributes Use large language models (LLMs), such as Transformers-based models, to extract entities, relations, and attributes.

[0050] Named Entity Recognition (NER): Automatically identify named entities such as dishes and diseases.

[0051] Relation extraction: extracting the relationship between entities from related corpus.

[0052] Attribute extraction: Collecting attribute information of a specific entity from different information sources.

[0053] Tools: Transformers-based models (Baichuan, Zhipu).

[0054] Reference Figure 5 As shown in the figure, the effect of automatically extracting entities, attributes and relationships from raw data is demonstrated. This process involves the following aspects: Named Entity Recognition (NER): Automatically identify key entities such as dish names and disease names in text.

[0055] Relationship extraction: Identifying relationships between entities, such as how a certain food affects the occurrence and development of a specific disease.

[0056] Attribute extraction: extracting specific features of each entity, such as the nutritional content of a dish, the typical symptoms of a certain disease, etc.

[0057] Visual display: The extraction results are presented in a graphical way to facilitate user understanding and analysis.

[0058] 5. Step 5: Knowledge Integration: Logical attribution and redundant filtering Entity linking: Entity disambiguation and coreference resolution are performed through similarity calculation, and entity reference items are linked to corresponding entities in the knowledge base.

[0059] Knowledge merging: handling entity conflicts at the data layer and the model layer, and merging external knowledge bases.

[0060] Tools: Use OpenRefine for data cleaning and ETL tools (such as Apache Nifi and Talend) for data merging.

[0061] Reference Figure 4 As shown in the figure, the process of handling conflicts and ambiguous entities is described to ensure the consistency and accuracy of the knowledge graph. The specific steps are as follows: Identify conflicts and ambiguities: Identify conflicting or ambiguous entities by comparing data from different sources.

[0062] Similarity calculation: For entities that appear to be identical, text similarity algorithms (such as edit distance, cosine similarity, etc.) are used for comparison.

[0063] Disambiguation and coreference resolution: Natural language processing (NLP) techniques, including but not limited to entity linking and semantic role labeling, are used to determine the true identity of entities.

[0064] Merge entities: Once multiple entities are confirmed to point to the same actual object, merge them into a unified entity.

[0065] Update the knowledge base: Finally, update the modified entity information to the knowledge graph to ensure the consistency of all relevant information.

[0066] 6. Step 6: Knowledge Calculation: 6.1. Ontology construction: The extracted entities, relations and attributes are constructed into a semantic ontology to classify the data into concept sets or frameworks.

[0067] 6.2. Knowledge reasoning and quality assessment: Use reasoning engines based on large language models (such as Apache Jena and HermiT) to handle missing relationship values ​​between subjects in the knowledge graph and complete further knowledge discovery; use consistency, accuracy and other evaluation indicators for quality assessment.

[0068] 6.3. Continuous update mechanism: including updates of the concept layer and data layer, checking the reliability and consistency of the data source, and ensuring the dynamic update and expansion of the knowledge graph.

[0069] 7. Step 7: Knowledge base storage: Use graph database Neo4J for persistent storage.

[0070] Ensure the integrity and queryability of knowledge.

[0071] Tools: Neo4J, Cayley, ArangoDB, etc.

[0072] So far, various embodiments of the present application have been described in detail. In order to avoid obscuring the concept of the present application, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed herein.

[0073] Finally, it should be noted that the above is only a preferred embodiment of the present application, and the aforementioned embodiments are only used to illustrate the technical solution of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that it is still possible to modify the technical solutions recorded in the aforementioned embodiments, or to make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not drive the essence of the corresponding technical solutions away from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An algorithm for constructing a disease-dish data knowledge graph, characterized in that: The following steps are involved: Determine the disease-dish knowledge modeling in the knowledge graph; Collect and store structured, unstructured and semi-structured raw food data and disease knowledge; Extracting entities, relationships, and entity attributes from the data using information extraction techniques; Knowledge fusion, which integrates the extracted information, including filtering redundant information and merging similar entities to obtain a unified knowledge representation; Perform knowledge computation, including quality assessment, missing value processing, and knowledge updating; The knowledge generated by the above process is stored in a graph database for easy query and use.

2. According to claim 1, the algorithm for constructing a disease-dish data knowledge graph is characterized in that: The knowledge modeling divides the knowledge graph into a pattern graph and a data graph, wherein the pattern graph is used to reflect the relationship between concepts, and the data graph is used to reflect the relationship between actual data.

3. The algorithm for constructing a disease-dish data knowledge graph according to claim 1, characterized in that: The information extraction technology includes using a large language model for entity recognition, relationship extraction and attribute extraction.

4. The algorithm for constructing a disease-dish data knowledge graph according to claim 3 is characterized in that: The large language model used is a model based on the Transformers architecture.

5. The algorithm for constructing a disease-dish data knowledge graph according to claim 1, characterized in that: The knowledge fusion further comprises: Entity linking: entity disambiguation and coreference resolution through similarity calculation; Knowledge merging: handling entity conflicts between data and schema layers from different sources and merging external knowledge bases with existing ones.

6. The algorithm for constructing a disease-dish data knowledge graph according to claim 1, characterized in that: It also includes creating semantic ontologies to structure entities, relationships, and attributes into concept sets or frameworks to classify and organize data.

7. The algorithm for constructing a disease-dish data knowledge graph according to claim 1, characterized in that: The knowledge calculation step also involves using an inference engine to perform data reasoning to supplement the missing relationship values ​​between subjects in the knowledge graph and complete further knowledge discovery.

8. The algorithm for constructing a disease-dish data knowledge graph according to claim 1, characterized in that: The graph database is selected from Neo4J, Cayley, and ArangoDB to ensure efficient storage and query capabilities and enhance system response speed.

9. An algorithm for constructing a disease-dish data knowledge graph according to any one of claims 1 to 9, characterized in that: A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for constructing a disease-dish data knowledge graph according to any of the preceding claims.

Images