Ophthalmologic knowledge graph construction method and device, equipment and medium

Abstract

The embodiment of the invention provides an ophthalmology knowledge graph construction method and device, equipment and a medium, and relates to the technical field of ophthalmology graph construction, and the method comprises the steps: obtaining an ophthalmology teaching material document, carrying out the theme division, obtaining at least two ophthalmology theme chapter blocks, carrying out the paragraph segmentation of each chapter block, and forming at least two ophthalmology paragraph blocks, then performing entity extraction on each paragraph block to obtain an initial ophthalmology entity list, performing semantic screening through a preset large language model in combination with semantic matching prompt words, paragraph topics and the initial entity list to obtain a target ophthalmology entity list, and standardizing a target entity name by using a term concept unique identifier mapping system to obtain a target ophthalmology entity list; according to the method, the target entity list is updated, finally, the basic ophthalmology knowledge graph is generated based on the updated target entity list, the basic ophthalmology knowledge graph is fused with the preset medical ontology library knowledge graph, and the extended ophthalmology knowledge graph is constructed, so that the efficiency and accuracy of obtaining information of related diseases are improved.

Description

Technical Field

[0001] The present application relates to the technical field of ophthalmic atlas construction, and in particular to an ophthalmic knowledge graph construction method, device, equipment and medium. Background Art

[0002] In related technologies, the same clinical entity in medical textbooks often has multiple expressions (for example, "vitreous hemorrhage" may be described as "vitreous hemorrhage", "intraocular blood accumulation", etc.), which makes entity normalization difficult, and the relationships generated by LLM are prone to logical contradictions (for example, "vitreous hemorrhage" is simultaneously associated with two different causes, "trauma" and "diabetes").

[0003] UMLS mainly relies on the mapping of existing terminology systems, has insufficient coverage of emerging fields (such as gene-drug interactions in precision medicine), lacks non-standardized terminology, and lacks systematic integration of non-standard expressions widely used in clinical practice, resulting in a semantic gap with real medical data. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art. To this end, this application proposes a method, device, equipment and medium for constructing an ophthalmological knowledge graph, aiming to improve the efficiency and accuracy of obtaining information on related diseases.

[0005] In a first aspect, an embodiment of the present application provides a method for constructing an ophthalmological knowledge graph, the method comprising: Access ophthalmology textbook documents; Dividing the ophthalmology textbook document by subject to obtain at least two ophthalmology subject chapter blocks; Performing paragraph segmentation on the ophthalmology subject chapter block to obtain at least two ophthalmology paragraph blocks; Performing entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes a plurality of candidate ophthalmology entities, and the candidate ophthalmology entities include entity names and concept unique identifiers; Performing entity semantic screening using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, each of the target ophthalmology entities being selected from the multiple candidate ophthalmology entities; performing terminology standardization on the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity to update the target ophthalmological entity list; A basic ophthalmology knowledge graph is generated according to the target ophthalmology entity list, and the basic ophthalmology knowledge graph is integrated with a preset medical ontology library knowledge graph to obtain an extended ophthalmology knowledge graph.

[0006] According to the technical solution of the embodiment of the present application, there are at least the following beneficial effects: the embodiment of the present invention first obtains ophthalmology textbook documents, which provides a good foundation for the subsequent topic division of ophthalmology textbook documents; by subject division and paragraph segmentation of the ophthalmology textbook documents, ophthalmology paragraph blocks are obtained, which facilitates subsequent entity extraction; through entity extraction technology, multiple candidate ophthalmology entities are identified from the ophthalmology paragraph blocks, and concept unique identifiers are assigned to them, and an initial ophthalmology entity list is constructed, which provides basic data for subsequent knowledge graph construction; using a preset large language model and semantic matching prompt words, the initial ophthalmology entity list is semantically screened to obtain a target ophthalmology entity list, ensuring the accuracy and relevance of the entities in the knowledge graph; according to a preset term concept unique identifier mapping system, the entity names of the target ophthalmology entities are terminologically standardized, eliminating the ambiguity and diversity of the terms, and improving the uniformity and standardization of the knowledge graph; based on the target ophthalmology entity list, a basic ophthalmology knowledge graph is generated, and it is integrated with the preset medical ontology library knowledge graph to obtain an extended ophthalmology knowledge graph, enriching the content and depth of the knowledge graph, thereby improving the efficiency and accuracy of obtaining information about related diseases.

[0007] According to some embodiments of the present application, the entity semantic screening is performed by using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a target ophthalmology entity list, including: Performing a relevance evaluation using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a relevance score for each candidate ophthalmology entity, and recording the largest relevance score as a maximum relevance score; Calculating the ratio of the relevance score of each candidate ophthalmological entity to the maximum relevance score to obtain a relevance score value; Determine the candidate ophthalmological entity whose correlation score is greater than a preset score threshold as the target ophthalmological entity; Entities that are not the target ophthalmology entities in the initial ophthalmology entity list are deleted to obtain the target ophthalmology entity list.

[0008] According to some embodiments of the present application, after performing entity semantic screening based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list using a preset large language model to obtain a target ophthalmology entity list, the method further includes: Constructing a synonym ring network based on the target ophthalmology entity list; Performing logic detection on the synonym ring network to obtain a logic detection result; If the logic detection result indicates that a conflicting entity exists, at least two causes associated with the conflicting entity are updated so that each target ophthalmic entity is associated with one cause.

[0009] According to some embodiments of the present application, the basic ophthalmology knowledge graph is integrated with a preset medical ontology knowledge graph to obtain an extended ophthalmology knowledge graph, including: Merging a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph to obtain a target entity node; wherein the first number is smaller than the second number; Generate a cross edge according to a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph; Merging the third number of cross edges, the fourth number of relationship edges in the basic ophthalmology knowledge graph, and the fifth number of relationship edges in the medical ontology knowledge graph to obtain target relationship edges; wherein the fourth number is smaller than the fifth number; The extended ophthalmology knowledge graph is constructed based on the target entity nodes and the target relationship edges.

[0010] According to some embodiments of the present application, the ophthalmology textbook text is divided into topics to obtain at least two ophthalmology topic chapter blocks, including: Obtaining a PDF catalogue analysis of the ophthalmology textbook document; Extracting the directory structure of the ophthalmology textbook based on the PDF directory parsing, and parsing the chapter information of the textbook; The ophthalmology textbook text is divided into topics according to the chapter information of the textbook to obtain at least two ophthalmology paragraph blocks.

[0011] According to some embodiments of the present application, the terminology standardization of the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity to update the target ophthalmological entity list includes: Based on a preset cross-term semantic association model, mapping the entity name of the target ophthalmology entity through the term concept unique identifier mapping system and the concept unique identifier of the target ophthalmology entity to obtain the entity name after term standardization; Inputting the entity name after the term standardization into a preset term library for verification to obtain a verification result; The target ophthalmology entity list is updated according to the verification result.

[0012] According to some embodiments of the present application, the cross-term semantic association model is obtained by the following steps, including: Extracting target semantic relationships between ophthalmic entities to be extracted in the ophthalmic textbook document based on a preset relationship network and the ophthalmic textbook document; Classifying and arranging the extracted target semantic relations to form a structured semantic relation representation; The cross-term semantic association model is constructed according to the semantic relationship representation.

[0013] In a second aspect, an embodiment of the present application provides an ophthalmological knowledge graph construction device, the device comprising: Get document module to get ophthalmology teaching materials documents; A subject division module, configured to divide the ophthalmology teaching material document into subjects to obtain at least two ophthalmology subject chapter blocks; A paragraph segmentation module, configured to segment the ophthalmology subject chapter block into paragraphs to obtain at least two ophthalmology paragraph blocks; an entity extraction module, configured to perform entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes a plurality of candidate ophthalmology entities, each candidate ophthalmology entity including an entity name and a concept unique identifier; a semantic screening module, configured to perform entity semantic screening based on a preset semantic matching prompt word, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list using a preset large language model to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, each of the target ophthalmology entities being selected from the multiple candidate ophthalmology entities; a standardization processing module, configured to perform terminology standardization on the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity, so as to update the target ophthalmological entity list; A graph fusion module is used to generate a basic ophthalmic knowledge graph based on the target ophthalmic entity list, and fuse the basic ophthalmic knowledge graph with the preset medical ontology library knowledge graph to obtain an extended ophthalmic knowledge graph.

[0014] In a third aspect, an embodiment of the present application provides a computer device, including: Memory, used to store programs; A processor is used to execute the program stored in the memory. When the processor executes the program stored in the memory, the processor is used to execute the method described in the first aspect above. In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to execute the method described in the first aspect above.

[0015] The solutions provided in the second to fourth aspects are used to implement or cooperate with the method provided in the first aspect, and therefore can achieve the same or corresponding beneficial effects as the first aspect, and will not be repeated here.

[0016] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0018] Figure 1 A schematic diagram of a process for constructing an ophthalmic knowledge graph according to an embodiment of the present application; Figure 2 A schematic diagram of a process for obtaining a target ophthalmic entity list provided in one embodiment of the present application; Figure 3 A flowchart of a subsequent method for obtaining a target ophthalmic entity list provided in one embodiment of the present application; Figure 4 A schematic diagram of a process for obtaining an expanded ophthalmological knowledge graph provided in one embodiment of the present application; Figure 5 A schematic diagram of a process for obtaining an ophthalmology topic chapter block provided in one embodiment of the present application; Figure 6 A schematic diagram of a process for updating a target ophthalmology entity list provided in one embodiment of the present application; Figure 7 A schematic diagram of a process for obtaining a cross-term semantic association model provided in one embodiment of the present application; Figure 8 A schematic diagram of an ophthalmic knowledge graph construction device provided in one embodiment of the present application; Figure 9 A schematic diagram of a computer device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical methods and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application. It should be noted that the term "multiple" (or multiple items) used in the description of the embodiments of this application means more than two. "greater than," "less than," and "exceed" are to be understood as excluding the number indicated, while "above," "below," and "within" are to be understood as inclusive. The use of terms such as "first" and "second" is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features. In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the situation where A exists alone, A and B exist at the same time, or B exists alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can mean: a exists alone, b exists alone, c exists alone, a and b exist at the same time, a and c exist at the same time, b and c exist at the same time, or a, b and c exist at the same time, wherein a, b, c can be single or multiple.

[0020] In the description of this application, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in this application based on the specific content of the technical solution.

[0021] In some cases, the same clinical entity in medical textbooks often has multiple expressions (e.g., "vitreous hemorrhage" may be described as "vitreous hemorrhage", "intraocular blood accumulation", etc.), which makes entity normalization difficult and the relationships generated by LLM are prone to logical contradictions (e.g., associating "vitreous hemorrhage" with two different causes, "trauma" and "diabetes").

[0022] UMLS mainly relies on the mapping of existing terminology systems, has insufficient coverage of emerging fields (such as gene-drug interactions in precision medicine), lacks non-standardized terminology, and lacks systematic integration of non-standard expressions widely used in clinical practice, resulting in a semantic gap with real medical data.

[0023] With the in-depth development of medical informatization, knowledge graphs have evolved from auxiliary tools to core reasoning engines for intelligent diagnosis and treatment systems. While the current mainstream solution relies on the Unified Medical Language System (UMLS) as the authoritative knowledge source, its limitations are becoming increasingly apparent. As the world's largest open-source biomedical knowledge graph, UMLS integrates over 200 terminologies, but it suffers from bottlenecks such as insufficient coverage of ophthalmological disease entities (e.g., "central serous chorioretinopathy" is not standardized), sparse semantic relationships (only basic hierarchical relationships are defined), and delayed update cycles (annual releases make it difficult to incorporate the latest diagnostic and treatment guidelines in a timely manner). For example, its coverage of drug-indication relationships is only 68%, and its ability to represent multimodal knowledge such as imaging features and genomic data is weak.

[0024] The UMLS knowledge base transplantation approach still has the following shortcomings: limited coverage. UMLS primarily relies on mapping existing terminology systems, and lacks coverage of emerging fields (such as gene-drug interactions in precision medicine). For example, the latest biomarkers related to tumor immunotherapy (such as dynamic monitoring of PD-L1 expression) may not be included in a timely manner. Non-standardized terminology is missing. Non-standard expressions widely used in clinical practice (such as local diagnostic aliases and descriptions in unstructured electronic medical records) lack systematic integration, resulting in a semantic gap with real medical data. Knowledge updates are lagging. UMLS is typically updated annually, which cannot promptly reflect rapidly evolving medical knowledge (such as the latest pathological mechanisms of COVID-19 sequelae). Studies have shown that approximately 23% of clinical decisions rely on research results published in the past two years, and UMLS has limited ability to capture such dynamic knowledge. Semantic relationship sparsity. Existing relationships are mainly hierarchical classifications (such as "is_a") and basic semantic types, lacking fine-grained contextual associations (such as "Drug A affects the prognosis of disease C by inhibiting pathway B"). This makes it difficult for UMLS-based reasoning models to capture complex clinical decision logic.

[0025] Based on the above situation, this application proposes a method, device, equipment and medium for constructing an ophthalmic knowledge graph, aiming to improve the efficiency and accuracy of obtaining information on related diseases.

[0026] To facilitate understanding of the solutions of the embodiments of the present application, the relevant concepts involved in the embodiments of the present application are introduced below.

[0027] PDF parsing tool: A software or library used to read, extract, and process the content of PDF documents. It can help users extract text, images, tables, directory structures, and other information from PDF files and convert them into other formats (such as text files, JSON, XML, etc.) for further analysis or processing.

[0028] Term concept unique identifier mapping system: a system that identifies and associates term concepts in different knowledge bases. It assigns a unique identifier to each term concept to ensure that terms from different sources can be accurately matched and associated semantically. It assigns a unique identifier to each term concept to ensure that terms from different sources can be accurately identified, maps synonyms and related terms in different glossaries to the same concept, supports cross-system term conversion and interoperability, establishes semantic relationships between terms, helps understand the meaning of terms in different contexts, supports data integration between different data sources, and improves the consistency and accuracy of information.

[0029] Cross-term semantic association model: A model used to identify and associate terms from different glossaries or knowledge bases, ensuring semantic consistency and relevance between these terms. Using semantic analysis techniques, it establishes semantic relationships between terms, supports cross-system term conversion and interoperability, identifies semantic relationships between different terms, including synonyms, antonyms, and related terms, calculates semantic similarity between terms, and quantifies the strength of semantic associations between terms. It also maps terms from different glossaries to unified concept identifiers, supports term standardization, understands the meaning of terms in different contexts, and ensures the accuracy of semantic associations.

[0030] SciBERT model: A pre-trained language model based on BERT, optimized specifically for scientific text. SciBERT uses a vocabulary (scivocab) designed specifically for scientific text, which can better handle the professional terms and expressions in scientific literature. SciBERT is mainly used for natural language processing tasks in scientific literature, including scientific literature classification and retrieval, automatic summary generation, key information extraction, scientific question-answering systems, and assisting in the writing of literature reviews.

[0031] The GPT-4 system is a large-scale, multimodal language model. It utilizes the fourth-generation Generative Pre-trained Transformer (GPT4) developed by OpenAI. It can process text and image inputs and generate text outputs. It supports mixed text and image inputs, understands image content, and generates relevant text outputs. It can handle longer text inputs (approximately 25,000 words) and possesses stronger reasoning and logic capabilities. It provides customized services based on user needs to meet specific tasks. Incorporating DALL·E technology, it can generate high-quality images from text descriptions and supports multiple style transfers. It employs advanced model compression techniques and inference algorithms to reduce computational complexity. It optimizes model behavior and reduces harmful outputs through adversarial testing and reinforcement learning with human feedback (RLHF).

[0032] Synonym Ring Network: A network structure used to represent and manage the relationships between synonyms and near-synonyms. It connects related words together through a ring-like connection method, facilitating quick search and processing of synonyms. Synonyms and near-synonyms are associated through a ring structure, facilitating quick search and processing. Semantic Similarity Calculation: Calculates the semantic similarity between words through the connection method of the ring network, supports data integration between different data sources, improves the consistency and accuracy of information, helps understand the meaning of words in different contexts, and ensures the accuracy of semantic associations.

[0033] The ophthalmic knowledge graph construction method provided in the embodiments of the present application can be applied to a terminal, a server, or software running on a terminal or a server. In some embodiments, the terminal can be a smartphone, a tablet computer, a laptop computer, a desktop computer, etc.; the server can be configured as an independent physical server, or as a server cluster or distributed system consisting of multiple physical servers, or as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms; the software can be an application that implements the ophthalmic knowledge graph construction method, etc., but is not limited to the above forms. Embodiments of the present application can be used in a wide variety of general-purpose or specialized computer system environments or configurations. For example, personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above, and the like. The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communications network. In distributed computing environments, program modules may be located in local and remote computer storage media, including storage devices. It should be noted that in various embodiments of the present invention, when data related to the characteristics of an object (e.g., a user) is processed based on its attribute information or attribute information sets, permission or consent from the corresponding object will be obtained. Furthermore, the collection, use, and processing of this data will comply with relevant laws and standards. In addition, when the embodiment of the present invention needs to obtain the attribute information of an object, it will obtain the separate permission or separate consent of the corresponding object through a pop-up window or jump to a confirmation page. After clearly obtaining the separate permission or separate consent of the corresponding object, it will obtain the relevant data of the object necessary to enable the normal operation of the embodiment of the present invention.

[0034] See also Figure 1 , Figure 1 A flowchart of a method for constructing an ophthalmic knowledge graph provided in one embodiment of the present application is provided. The method for constructing an ophthalmic knowledge graph provided in one embodiment of the present application includes but is not limited to steps S110 to S170, and each step is introduced in sequence below. Step S110: Obtain ophthalmology teaching material documents.

[0035] Specifically, ophthalmology teaching materials can include online databases, electronic teaching materials published by academic institutions, and in-house training materials of hospitals. Ophthalmology teaching materials can be in formats such as PDF, Word, and HTML.

[0036] Step S120: Divide the ophthalmology teaching material document into topics to obtain at least two ophthalmology topic chapter blocks.

[0037] Specifically, ophthalmology textbooks typically have multiple topical ophthalmology chapter blocks. Topics can include eye diseases (cataracts, glaucoma), treatment techniques (laser surgery), or anatomical structures (retina, lens). Each topical ophthalmology chapter block describes the ophthalmology knowledge associated with that topic. For example, an ophthalmology chapter block on cataracts would describe the definition and causes of cataracts.

[0038] Step S130: segment the ophthalmology subject chapter block into paragraphs to obtain at least two ophthalmology paragraph blocks.

[0039] Specifically, an ophthalmology section block includes multiple paragraphs. Each paragraph describes a portion of ophthalmology knowledge related to the topic. For example, an ophthalmology section block on the topic of cataracts includes an ophthalmology paragraph block corresponding to the definition of cataracts and an ophthalmology paragraph block corresponding to the causes of cataracts.

[0040] Step S140 : performing entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes a plurality of candidate ophthalmology entities, and the candidate ophthalmology entities include entity names and concept unique identifiers.

[0041] In one embodiment, the SciBERT model is used to perform entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list. The entities may be disease names, symptoms, treatment methods, drug names, anatomical structures, etc.

[0042] Step S150: Entity semantics screening is performed using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, and each target ophthalmology entity comes from multiple candidate ophthalmology entities.

[0043] In one embodiment, a preset GPT large language model can be used, fine-tuned with domain knowledge, and the semantic matching prompt word rule is to screen entities related to the current topic as an ophthalmologist. For example, if the preset semantic matching prompt word is "Please filter entities related to 'cataract' from the following list," a list of candidate entities is input to obtain a list of target entities such as "cataract" and "lens opacity."

[0044] Step S160 : performing terminology standardization on the entity name of the target ophthalmological entity according to the preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity to update the target ophthalmological entity list.

[0045] In one implementation, a term-concept unique identifier mapping system is obtained through the following steps, including: obtaining the MRCONSO.RRF file of UMLS; preprocessing the MRCONSO.RRF file to obtain a preprocessed MRCONSO.RRF file; extracting term information based on the preprocessed MRCONSO.RRF file; associating the term information with the concept identifier in UMLS to obtain a term-concept unique identifier mapping system.

[0046] Step S170: Generate a basic ophthalmology knowledge graph based on the target ophthalmology entity list, and fuse the basic ophthalmology knowledge graph with the preset medical ontology library knowledge graph to obtain an extended ophthalmology knowledge graph.

[0047] It should be noted that the Neo4j subgraph embedding alignment technology can be used to merge the basic ophthalmology knowledge graph with the preset medical ontology knowledge graph.

[0048] In the embodiment of the present application, after obtaining an ophthalmology textbook document, the subject division is first performed to obtain at least two ophthalmology subject chapter blocks. Then, each chapter block is segmented into paragraphs to form at least two ophthalmology paragraph blocks. Then, entity extraction is performed on each paragraph block to obtain an initial ophthalmology entity list containing entity names and concept unique identifiers. Semantic screening is performed through a preset large language model in combination with semantic matching prompt words, paragraph topics and the initial entity list to obtain a target ophthalmology entity list. The target entity names are standardized using a term-concept unique identifier mapping system, and the target entity list is updated. Finally, a basic ophthalmology knowledge graph is generated based on the updated target entity list, and is integrated with a preset medical ontology library knowledge graph to construct an extended ophthalmology knowledge graph, thereby improving the efficiency and accuracy of obtaining information on related diseases.

[0049] See also Figure 2 , Figure 2 A flowchart of obtaining a target ophthalmological entity list is provided for one embodiment of the present application. Regarding the above step S150, including but not limited to steps S210 to S240, each step is introduced in sequence below. Step S210: Perform relevance evaluation based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list through a preset large language model to obtain a relevance score for each candidate ophthalmology entity, and record the maximum relevance score as the maximum relevance score.

[0050] It should be noted that through the preset large language model, combined with semantic matching prompt words, input text and a list of entities to be screened, the input text is a description of an ophthalmic clinical topic, and the relevance score of each candidate ophthalmic entity is obtained. The value of the relevance score is greater than or equal to 0 and less than or equal to 1. The obtained relevance scores are compared, and the largest relevance score is recorded as the maximum relevance score.

[0051] Step S220 : Calculate the ratio of the relevance score of each candidate ophthalmological entity to the maximum relevance score to obtain a relevance score value.

[0052] It should be noted that the relevance score values are normalized, and the relevance score of each candidate entity is divided by the maximum relevance score to obtain the relevance score value.

[0053] Step S230: Determine the candidate ophthalmological entity whose relevance score is greater than a preset score threshold as the target ophthalmological entity.

[0054] It should be noted that the preset scoring threshold is used to screen target entities. For example, the threshold is set to 0.7, and the relevance score value is compared with the preset scoring threshold. Entities with relevance scores greater than or equal to the threshold are determined as target entities.

[0055] Step S240: Delete the entities in the initial ophthalmology entity list that are not target ophthalmology entities to obtain a target ophthalmology entity list.

[0056] It should be noted that entities with a relevance score value less than the threshold are not target ophthalmology entities and are deleted from the initial entity list. The target ophthalmology entities are retained to obtain the target ophthalmology entity list.

[0057] In one embodiment, the topic is common symptoms of cataracts, and the candidate entity list is: ["decreased vision", "eye pain", "headache", "nausea", "vomiting", "lens opacity", "glaucoma"]. The GPT model is used to evaluate the relevance of each entity to obtain a score of 0.95 for decreased vision, 0.90 for lens opacity, 0.75 for eye pain, 0.40 for headache, 0.30 for nausea, 0.25 for vomiting, and 0.10 for glaucoma. The maximum relevance score is 0.95. The relevance score of each candidate entity is divided by the maximum relevance score to obtain a relevance score value. The relevance score value of decreased vision is 1.00, the relevance score value of lens opacity is approximately 0.95, the relevance score value of eye pain is approximately 0.79, the relevance score value of headache is approximately 0.42, the relevance score value of nausea is approximately 0.32, the relevance score value of vomiting is approximately 0.26, and the relevance score value of glaucoma is approximately 0.10. When the preset score threshold is 0.7, the target ophthalmology entities include decreased vision (1.00), lens opacity (0.95), and eye pain (0.79), and the target entity list is ["decreased vision", "lens opacity", "eye pain"] See also Figure 3 , Figure 3 A flowchart of a subsequent method for obtaining a target ophthalmic entity list provided in one embodiment of the present application. After the above step S150, including but not limited to steps S310 to S330, each step is introduced in sequence below. Step S310: construct a synonym ring network based on the target ophthalmology entity list.

[0058] It should be noted that the synonym ring network is a graph structure used to represent the relationship between target ophthalmology entities and their synonyms. Each node represents a target ophthalmology entity, and edges represent synonym relationships between entities. Natural language processing techniques (such as word vector similarity calculation and dictionary matching) are used to identify synonyms. Entities and their synonyms are represented in a graph structure, ensuring that each entity is connected to its synonym by an edge.

[0059] Step S320: Perform logic detection on the synonym ring network to obtain a logic detection result.

[0060] It should be noted that there are logical conflicts in the synonym ring network, such as the same entity being associated with multiple unrelated synonyms; inconsistent synonym relationships (such as A is a synonym of B, but B is not a synonym of A); and contradictions in the ring structure (such as A → B → C → A, but A and C have no direct relationship).

[0061] Step S330: If the logic detection result indicates that a conflicting entity exists, at least two causes associated with the conflicting entity are updated so that each target ophthalmological entity is associated with one cause.

[0062] It should be noted that when conflicting entities are found during logical detection, the causes associated with the conflicting entities are updated based on medical knowledge and the use of a pre-trained large language model to assist in verification, ensuring that each target ophthalmic entity is associated with an accurate cause.

[0063] See also Figure 4 , Figure 4 A schematic diagram of a process for obtaining an expanded ophthalmology knowledge graph according to an embodiment of the present application is provided. Regarding the above-mentioned step S170, including but not limited to steps S410 to S440, each step is introduced in sequence below. Step S410: Merge a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph to obtain target entity nodes; wherein the first number is smaller than the second number; Step S420: generating cross edges based on a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph; Step S430: Merge the third number of cross edges, the fourth number of relationship edges in the basic ophthalmology knowledge graph, and the fifth number of relationship edges in the medical ontology knowledge graph to obtain target relationship edges; wherein the fourth number is smaller than the fifth number; Step S440: Construct an extended ophthalmology knowledge graph based on the target entity nodes and target relationship edges.

[0064] In one embodiment, 566 entity nodes in the basic ophthalmology knowledge graph and 26,815 entity nodes in the medical ontology knowledge graph are merged to obtain target entity nodes, cross edges are generated based on the 566 entity nodes in the basic ophthalmology knowledge graph and the 26,815 entity nodes in the medical ontology knowledge graph, and the cross edges, 890 relationship edges in the basic ophthalmology knowledge graph, and 35,383 relationship edges in the medical ontology knowledge graph are merged to obtain target relationship edges.

[0065] In one embodiment, 566 entity nodes were extracted from the basic ophthalmology knowledge graph. These entities can be disease names, symptoms, treatment methods, and drug names, such as "glaucoma," "blurred vision," "increased intraocular pressure," and "intraocular pressure-lowering drugs." 890 relationship edges were extracted from the basic ophthalmology knowledge graph. These relationships can be "cause," "symptom," and "treatment method." For example, there is a "cause" relationship between "glaucoma" and "increased intraocular pressure," and a "symptom" relationship between "blurred vision" and "glaucoma."

[0066] See also Figure 5 , Figure 5 This is a flow chart of obtaining an ophthalmology theme chapter block according to one embodiment of the present application. Regarding the above step S130, including but not limited to steps S510 to S530, each step is introduced in sequence below. Step S510: Obtain the PDF directory of the ophthalmology teaching material document.

[0067] Step S520: extract the directory structure of the ophthalmology textbook based on the PDF directory parsing, and parse out the chapter information of the textbook.

[0068] Step S530: Divide the ophthalmology textbook text into topics according to the chapter information of the textbook to obtain at least two ophthalmology paragraph blocks.

[0069] In one embodiment, a PDF parsing tool is used to read a PDF document of an ophthalmology textbook. By parsing the directory portion of the PDF document, the chapter titles and corresponding page number information of the textbook are extracted, and the extracted directory structure is stored for subsequent processing. Then, the chapter titles and corresponding page number information are extracted from the parsed directory structure, and a hierarchical structure of the chapters is constructed to clarify the parent-child relationship between the chapters. Based on the page number information of the chapters, the text content of the PDF document is divided into chapters, and the text content of each chapter is further divided into paragraph blocks. The subject is marked for each paragraph block to facilitate the subsequent construction of a knowledge graph. Through the above steps, the complete directory structure of the textbook, including the chapter titles and page number information, is extracted and stored, which is convenient for subsequent text processing and clarifies the directory structure. At the same time, structured text data is provided for the subsequent construction of a knowledge graph to facilitate the extraction of entities and relationships.

[0070] In one embodiment, the chapter titles and corresponding page number information are extracted from the parsed directory structure, for example: Chapter 1: Ophthalmic Anatomy (Page 10) 1.1 Eyeball Structure (Page 10) 1.2 Ocular appendages (page 15) Chapter 2: Eye Diseases (Page 25) 2.1 Glaucoma (page 25) 2.2 Cataract (Page 30) Chapter 1 (parent chapter) contains two sub-chapters, 1.1 and 1.2. Chapter 2 (parent chapter) contains two sub-chapters, 2.1 and 2.2. The text content of Chapter 1 is from pages 10 to 15, and the text content of Chapter 2 is from pages 25 to 30. The text of Chapter 1 can be divided into multiple paragraph blocks, such as "Anatomical characteristics of the eye structure", "Functions of ocular appendages", etc. The text of Chapter 2 can be divided into "Causes and symptoms of glaucoma", "Treatments of cataracts", etc. Paragraph block 1: The theme is "Anatomical characteristics of the eye structure", Paragraph block 2: The theme is "Causes and symptoms of glaucoma".

[0071] The results after processing are as follows: Paragraph block for Chapter 1: Section 1: Anatomical features of the eyeball (Topic: Eyeball structure) Block 2: Functions of the ocular appendages (Topic: Ocular appendages) Paragraph block for Chapter 2: Section 3: Causes and Symptoms of Glaucoma (Topic: Glaucoma) Section 4: Treatment of Cataracts (Topic: Cataracts) See also Figure 6 , Figure 6 A flowchart of updating a target ophthalmology entity list according to an embodiment of the present application is provided. Regarding the above step S160, including but not limited to steps S610 to S630, each step is introduced in sequence below. Step S610: Based on the preset cross-term semantic association model, the entity name of the target ophthalmology entity is mapped through the term concept unique identifier mapping system and the concept unique identifier of the target ophthalmology entity to obtain the entity name after term standardization.

[0072] Step S620: Input the entity name after terminology standardization into a preset terminology database for verification to obtain a verification result.

[0073] Step S630: Update the target ophthalmology entity list according to the verification result.

[0074] In one embodiment, in the ophthalmic knowledge graph, "glaucoma" is mapped to "Glaucoma" in UMLS as a unique identifier to ensure the semantic consistency of "increased intraocular pressure" and "increased intraocular pressure". A cross-term semantic association model is used to associate "glaucoma" with "increased intraocular pressure". The semantic similarity between "glaucoma" and "increased intraocular pressure" is calculated through the model, "increased intraocular pressure" is mapped to the standardized name "increased intraocular pressure", and "glaucoma" is mapped to the standardized name "Glaucoma" in UMLS. "Glaucoma" is entered into the UMLS terminology library for verification to confirm its existence and correctness, and the semantics of "increased intraocular pressure" in UMLS are verified to be correct. The verification results show that "Glaucoma" exists in UMLS and is semantically correct. The verification results show that "increased intraocular pressure" is semantically correct in UMLS, and "glaucoma" is updated to "Glaucoma", and "increased intraocular pressure" is updated to "increased intraocular pressure".

[0075] Assume that the target ophthalmology entity list contains the following terms: glaucoma, increased intraocular pressure, cataract.

[0076] By implementing the above steps, the following results can be obtained: (1) Terminology standardization: "Glaucoma" is mapped to "Glaucoma", "intraocular pressure" is mapped to "intraocular pressure increase", and "cataract" is mapped to "Cataract"; (2) Terminology validation: The validation results show that “Glaucoma” exists in UMLS and is semantically correct. The validation results show that “intraocular pressure increase” exists in UMLS and is semantically correct. The validation results show that “Cataract” exists in UMLS and is semantically correct. (3) Update the target ophthalmology entity list to obtain the updated entity list: Glaucoma, increased intraocular pressure, and Cataract.

[0077] See also Figure 7 , Figure 7 A flowchart of obtaining a cross-term semantic association model is provided for one embodiment of the present application. The process of obtaining a cross-term semantic association model includes but is not limited to steps S710 to S730, and each step is described in sequence below. Step S710 : extracting target semantic relationships between ophthalmology entities to be extracted in the ophthalmology textbook document based on a preset relationship network and the ophthalmology textbook document.

[0078] It should be noted that the preset relationship network is used to describe the possible relationships between ophthalmic entities. Entity recognition technology is used to extract ophthalmic entities from ophthalmic textbook documents. Relation extraction technology is used to identify the semantic relationships between entities. Semantic analysis technology is used to understand the semantic associations between entities.

[0079] Step S720: Classify and organize the extracted target semantic relationships to form a structured semantic relationship representation.

[0080] It should be noted that the extracted semantic relations are classified according to predefined categories, the classified semantic relations are organized into a structured representation, and the structured semantic relations are stored in a database or knowledge graph to facilitate subsequent processing and query.

[0081] Step S730: construct a cross-term semantic association model based on the semantic relationship representation.

[0082] It should be noted that based on the structured semantic relationship representation, a cross-term semantic association model is constructed. By constructing a cross-term semantic association model, the semantic association between different terms is realized, and the uniformity and comparability of terms are improved. In one embodiment, relationships in a relationship network may include "cause," "symptoms," and "treatment methods." Entities such as "glaucoma," "increased intraocular pressure," and "blurred vision" are identified. Relationship extraction techniques are used to identify the "cause" relationship between "glaucoma" and "increased intraocular pressure." Semantic analysis techniques are used to analyze the "symptom" relationship between "blurred vision" and "glaucoma." "Cause," "symptom," and "treatment methods" are used as primary categories, and a triplet (entity 1, relationship, entity 2) is used to represent the semantic relationship. For example, for ("glaucoma," "cause," and "increased intraocular pressure"), the semantic similarity between "increased intraocular pressure" and "increased intraocular pressure" is calculated, and "increased intraocular pressure" is mapped to "increased intraocular pressure."

[0083] In one embodiment, the directory structure of the ophthalmology textbook is first extracted based on PDF directory parsing, and the chapter information of the textbook is parsed, which provides a good foundation for the subsequent topic division of the ophthalmology textbook text; the ophthalmology entity list is processed through the SciBERT model, GPT-4 system, and synonym ring network to ensure the accuracy and consistency of the ophthalmology entity list, which facilitates subsequent graph fusion; the basic ophthalmology knowledge graph is fused with the preset medical ontology library knowledge graph based on Neo4j subgraph embedding through the graph structure merging algorithm to eliminate the redundancy and ambiguity of the entities, thereby improving the efficiency and accuracy of obtaining information on related diseases.

[0084] The present application also provides an ophthalmology knowledge graph construction device, such as Figure 8 As shown, the device includes: Get document module to get ophthalmology teaching materials documents; A subject division module is used to divide the ophthalmology textbook document into subjects to obtain at least two ophthalmology subject chapter blocks; A paragraph segmentation module, configured to segment the ophthalmology subject chapter block into paragraphs to obtain at least two ophthalmology paragraph blocks; An entity extraction module is used to extract entities from each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes multiple candidate ophthalmology entities, and the candidate ophthalmology entities include entity names and concept unique identifiers; a semantic screening module, configured to perform entity semantic screening using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, each target ophthalmology entity being selected from multiple candidate ophthalmology entities; a standardization processing module, configured to perform terminology standardization on the entity name of the target ophthalmology entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmology entity, so as to update the target ophthalmology entity list; The graph fusion module is used to generate a basic ophthalmic knowledge graph based on the target ophthalmic entity list, and fuse the basic ophthalmic knowledge graph with the preset medical ontology library knowledge graph to obtain an extended ophthalmic knowledge graph.

[0085] It should be noted that the information interaction, execution process and other contents between the above modules are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0086] Figure 9 The computer device 900 provided in an embodiment of the present application is shown. The computer device 900 may be a server or a terminal. The internal structure of the computer device 900 includes but is not limited to: Memory 910, used for storing programs; The processor 920 is used to execute the program stored in the memory 910. When the processor 920 executes the program stored in the memory 910, the processor 920 is used to execute the above-mentioned ophthalmic knowledge graph construction method. The processor 920 and the memory 910 may be connected via a bus or other means. The memory 910, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs, such as the ophthalmic knowledge graph construction method described in any embodiment of this application. The processor 920 implements the above-described ophthalmic knowledge graph construction method by executing the non-transitory software program and instructions stored in the memory 910. The memory 910 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store and execute the above-mentioned ophthalmic knowledge graph construction method. In addition, the memory 910 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 910 may optionally include a memory remotely located relative to the processor 920, and these remote memories may be connected to the processor 920 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof. The non-transient software programs and instructions required to implement the above-mentioned ophthalmic knowledge graph construction method are stored in the memory 910. When executed by one or more processors 920, the ophthalmic knowledge graph construction method provided in any embodiment of the present application is executed. An embodiment of the present application also provides a computer-readable storage medium storing computer-executable instructions, which are used to execute the above-mentioned ophthalmic knowledge graph construction method. In one embodiment, the storage medium stores computer-executable instructions, which are executed by one or more control processors, for example, by one or more processors 920 in the above-mentioned computer device 900, so that the above-mentioned one or more processors 920 can execute the ophthalmic knowledge graph construction method provided in any embodiment of the present application. The embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment. Those skilled in the art will appreciate that all or some of the steps and systems disclosed above can be implemented as software, firmware, hardware, or any suitable combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on computer-readable media, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is well known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVDs) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. Furthermore, as is well known to those skilled in the art, communication media typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media. The above is a detailed description of the preferred implementation of the present application, but the present application is not limited to the above implementation. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present application. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A method for constructing an ophthalmological knowledge graph, characterized in that: The method comprises: Access ophthalmology textbook documents; Dividing the ophthalmology textbook document by subject to obtain at least two ophthalmology subject chapter blocks; Performing paragraph segmentation on the ophthalmology subject chapter block to obtain at least two ophthalmology paragraph blocks; Performing entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes a plurality of candidate ophthalmology entities, and the candidate ophthalmology entities include entity names and concept unique identifiers; Performing entity semantic screening using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, each of the target ophthalmology entities being selected from the multiple candidate ophthalmology entities; performing terminology standardization on the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity to update the target ophthalmological entity list; A basic ophthalmology knowledge graph is generated according to the target ophthalmology entity list, and the basic ophthalmology knowledge graph is integrated with a preset medical ontology library knowledge graph to obtain an extended ophthalmology knowledge graph.

2. The method according to claim 1, characterized in that The target ophthalmology entity list is obtained by performing entity semantic screening based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list using a preset large language model, including: Performing a relevance evaluation using a preset large language model based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list to obtain a relevance score for each candidate ophthalmology entity, and recording the largest relevance score as a maximum relevance score; Calculating the ratio of the relevance score of each candidate ophthalmological entity to the maximum relevance score to obtain a relevance score value; Determine the candidate ophthalmological entity whose correlation score is greater than a preset score threshold as the target ophthalmological entity; Entities that are not the target ophthalmology entities in the initial ophthalmology entity list are deleted to obtain the target ophthalmology entity list.

3. The method according to claim 1, characterized in that After performing entity semantic screening based on preset semantic matching prompt words, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list using a preset large language model to obtain a target ophthalmology entity list, the method further includes: Constructing a synonym ring network based on the target ophthalmology entity list; Performing logic detection on the synonym ring network to obtain a logic detection result; If the logic detection result indicates that a conflicting entity exists, at least two causes associated with the conflicting entity are updated so that each target ophthalmic entity is associated with one cause.

4. The method according to any one of claims 1 to 3, characterized in that The basic ophthalmology knowledge graph is integrated with the preset medical ontology knowledge graph to obtain an extended ophthalmology knowledge graph, including: Merging a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph to obtain a target entity node; wherein the first number is smaller than the second number; Generate a cross edge according to a first number of entity nodes in the basic ophthalmology knowledge graph and a second number of entity nodes in the medical ontology knowledge graph; Merging the third number of cross edges, the fourth number of relationship edges in the basic ophthalmology knowledge graph, and the fifth number of relationship edges in the medical ontology knowledge graph to obtain target relationship edges; wherein the fourth number is smaller than the fifth number; The extended ophthalmology knowledge graph is constructed based on the target entity nodes and the target relationship edges.

5. The method according to any one of claims 1 to 3, characterized in that The ophthalmology textbook text is divided into topics to obtain at least two ophthalmology topic chapter blocks, including: Obtaining a PDF catalogue analysis of the ophthalmology textbook document; Extracting the directory structure of the ophthalmology textbook based on the PDF directory parsing, and parsing the chapter information of the textbook; The ophthalmology textbook text is divided into topics according to the chapter information of the textbook to obtain at least two ophthalmology paragraph blocks.

6. The method according to any one of claims 1 to 3, characterized in that The terminology standardization of the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity to update the target ophthalmological entity list includes: Based on a preset cross-term semantic association model, mapping the entity name of the target ophthalmology entity through the term concept unique identifier mapping system and the concept unique identifier of the target ophthalmology entity to obtain the entity name after term standardization; Inputting the entity name after the term standardization into a preset term library for verification to obtain a verification result; The target ophthalmology entity list is updated according to the verification result.

7. The method according to claim 6, characterized in that The cross-term semantic association model is obtained by the following steps, including: Extracting target semantic relationships between ophthalmic entities to be extracted in the ophthalmic textbook document based on a preset relationship network and the ophthalmic textbook document; Classifying and arranging the extracted target semantic relations to form a structured semantic relation representation; The cross-term semantic association model is constructed according to the semantic relationship representation.

8. An ophthalmology knowledge graph construction device, characterized in that: The device comprises: Get document module to get ophthalmology teaching materials documents; A subject division module, configured to divide the ophthalmology teaching material document into subjects to obtain at least two ophthalmology subject chapter blocks; A paragraph segmentation module, configured to segment the ophthalmology subject chapter block into paragraphs to obtain at least two ophthalmology paragraph blocks; an entity extraction module, configured to perform entity extraction on each ophthalmology paragraph block to obtain an initial ophthalmology entity list, wherein the initial ophthalmology entity list includes a plurality of candidate ophthalmology entities, each candidate ophthalmology entity including an entity name and a concept unique identifier; a semantic screening module, configured to perform entity semantic screening based on a preset semantic matching prompt word, the topic corresponding to the ophthalmology paragraph block, and the initial ophthalmology entity list using a preset large language model to obtain a target ophthalmology entity list, wherein the target ophthalmology entity list includes multiple target ophthalmology entities, each of the target ophthalmology entities being selected from the multiple candidate ophthalmology entities; a standardization processing module, configured to perform terminology standardization on the entity name of the target ophthalmological entity according to a preset term concept unique identifier mapping system and the concept unique identifier of the target ophthalmological entity, so as to update the target ophthalmological entity list; A graph fusion module is used to generate a basic ophthalmic knowledge graph based on the target ophthalmic entity list, and fuse the basic ophthalmic knowledge graph with the preset medical ontology library knowledge graph to obtain an extended ophthalmic knowledge graph.

9. A computer device, characterized in that: include: Memory, used to store programs; A processor, configured to execute the program stored in the memory. When the processor executes the program stored in the memory, the processor is configured to execute: the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Computer executable instructions are stored. The execution instruction is used to execute: the method according to any one of claims 1 to 7.

Images