Knowledge base construction method, system and equipment based on large model

Abstract

The invention relates to a knowledge base construction method, system and equipment based on a large model, and relates to the technical field of artificial intelligence, by adopting the method, system and equipment, a main body and a preset association attribute in original corpus data are extracted through the large model, so that candidate knowledge units associated with the main body and the preset association attribute are obtained through the large model, namely, the candidate knowledge units of the main body and the preset association attribute are obtained. Through flattening redundant storage and knowledge embedding driven by a large model, one-key knowledge unit construction is realized, and dependence of a traditional scheme on a complex association structure is eliminated.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, system and device for constructing a knowledge base based on a large model. Background Art

[0002] Traditional knowledge base and knowledge graph technologies have been widely used for the storage and retrieval of structured knowledge, particularly in the fields of medical knowledge management and clinical decision support. However, existing technologies still have significant shortcomings in terms of knowledge relevance expression, dynamic update mechanisms, and complex query efficiency, which restricts their application in real-time clinical scenarios.

[0003] Traditional knowledge base technology usually uses a knowledge base system based on vector retrieval, which has the following key problems: (1) Lack of implicit associations: The causal chain between knowledge units cannot be explicitly expressed. For example, when a user queries the risks associated with propofol, the system can only return discrete entries such as "propofol pharmacological properties" and "hypotension management guidelines." The complete clinical logic of "propofol → hypotension risk → need to monitor blood pressure" needs to be manually pieced together, resulting in delayed decision-making.

[0004] (2) Dynamic updates are expensive: New knowledge requires a complete reconstruction of the vector space. For example, when a new drug is launched, not only do the embedding vectors of all related drugs need to be recalculated, but the weight parameters of the Rerank model also need to be adjusted to adapt to the characteristics of the new drug. Each update consumes thousands of GPU computing hours.

[0005] (3) Insufficient support for complex queries: Multiple condition combination queries are prone to information omission. In a typical scenario, for example, when querying "anesthesia plan for spinal surgery in patients with renal insufficiency", the system may miss key contraindication information (such as drugs that are not explicitly marked as "renal impairment" but are actually prohibited) because it cannot penetrate the three-layer logic of "renal insufficiency → drug metabolism contraindications → special requirements for spinal surgery".

[0006] Although knowledge graph technology can express the relationship between entities, its architecture has essential bottlenecks: (1) Storage and maintenance costs surge: When using "entity-edge-entity" triples for storage, the number of edges explodes in a combinatorial manner. Actual deployments show that for each new drug entity, multiple relationship edges (such as interactions, contraindications, etc.) need to be created, resulting in increased storage overhead.

[0007] (2) Query efficiency deterioration: The response time of multi-hop queries increases exponentially with depth. Tests show that for queries such as "find solutions containing propofol that are contraindicated for renal failure", when the associated path exceeds 3 hops, the response delay exceeds the clinically acceptable threshold of 2 seconds.

[0008] (3) Weak dynamic expansion capabilities: Adding new relationship types requires manual reconstruction of the graph. For example, when introducing real-world drug side effect data, the graph schema must be expanded after downtime, which is time-consuming and cannot meet the needs of rapid knowledge iteration in public health emergencies (such as updating therapeutic drugs).

[0009] Therefore, the existing technology has a dilemma of being highly dependent on complex association structures and being unable to achieve both expressiveness and efficiency. Summary of the Invention

[0010] In view of this, the purpose of the present invention is to provide a knowledge base construction method, system and equipment based on a large model to overcome the current problems of strong dependence on complex association structures and the inability to achieve both expressiveness and efficiency.

[0011] To achieve the above objectives, the present invention adopts the following technical solutions: In one aspect, a method for constructing a knowledge base based on a large model includes: Obtaining original corpus data of the target domain and annotating the original corpus data with entity types using a term library; the original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents; Extract the subject and its preset associated attributes from the original corpus data through the large model; Based on any corpus segment, according to the subject and preset association attributes, the embedded association information construction control instruction of the corpus segment is determined, and the embedded association information construction control instruction is input into the large model to obtain candidate knowledge units directly associated with the association attribute field and the subject field; Based on entity type and semantic consistency, the candidate knowledge units are initially detected for conflict. When the candidate knowledge units conflict, the conflict strategy is followed to obtain a knowledge unit set; wherein the knowledge unit set includes different non-conflicting knowledge units; The fields of the knowledge unit are structured and stored in JSON format; the fields include: the main body of current knowledge, directly related knowledge, a list of contraindications, a complication management plan, a list of equipment required for surgery, and version control information; the preset fields of the knowledge unit are vectorized and stored in the database to obtain a knowledge base.

[0012] Optionally, also include: The associated attribute fields of the knowledge units of the same subject are expanded. When the updated corpus of any associated attribute field is obtained, the knowledge units related to the updated corpus are updated.

[0013] Optionally, also include: In response to the assignment instruction, weights are assigned to the preset corpus; When the updated corpus conflicts with the corresponding content in the knowledge unit, the knowledge unit is updated according to the content with a higher weight.

[0014] Optionally, also include: Store the old and new versions of the corpus in parallel and timestamp them; or In response to the field priority setting instruction, the priority of the knowledge unit associated with the field is updated.

[0015] Optionally, also include: When there is duplicate information in any knowledge unit, deduplication is performed.

[0016] Optionally, also include: Record all field changes to build a version history table.

[0017] Optionally, also include: In response to natural language or fuzzy keywords input by the user, the relevant knowledge units are located through pre-stored entity vector indexes; By screening the structured fields, a complete knowledge unit is determined among the related knowledge units.

[0018] Optionally, also include: When the number of complete knowledge units is greater than 1, the similarities between different complete knowledge units are determined, and the complete knowledge unit with high credibility is determined as the target knowledge unit and returned to the user; or, When different versions of content exist in a complete knowledge unit, the content with a higher weight is determined as the content of the target knowledge unit and returned to the user.

[0019] In another aspect, a knowledge base construction system based on a large model includes: The input layer is used to obtain the original corpus data of the target domain and annotate the entity types of the original corpus data using the term library; the original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents; The processing layer is used to extract the subject and its preset associated attributes from the original corpus data through the large model; based on any corpus segment, determine the embedded associated information construction control instruction of the corpus segment according to the subject and preset associated attributes, and input the embedded associated information construction control instruction into the large model to obtain candidate knowledge units directly associated with the associated attribute field and the subject field; based on the entity type and semantic consistency, perform initial conflict detection on the candidate knowledge units, and when there is a conflict between the candidate knowledge units, execute according to the conflict strategy to obtain a knowledge unit set; wherein, the knowledge unit set includes different non-conflicting knowledge units; The storage layer is used to store the fields of the knowledge unit in a structured manner in JSON format; the fields include: the main body of the current knowledge, directly related knowledge, a list of contraindications, a complication management plan, a list of equipment required for surgery, and version control information; the preset fields of the knowledge unit are vectorized and stored in the database to obtain a knowledge base.

[0020] Optionally, a device for constructing a knowledge base based on a large model includes a processor and a memory, wherein the processor is connected to the memory: The processor is configured to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute any one of the above-mentioned methods for constructing a knowledge base based on a large model.

[0021] The technical solution provided by the embodiment of the present invention has at least the following beneficial effects: By adopting the technical solution provided by the embodiment of the present invention, the subject and preset associated attributes in the original corpus data are extracted through a large model, so that the candidate knowledge units associated with the subject and the preset associated attributes are obtained through the large model. That is, through flattened redundant storage and knowledge embedding driven by the large model, one-click knowledge unit construction is achieved, eliminating the traditional solution's dependence on complex associated structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of a flow chart of a method for constructing a knowledge base based on a large model provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a knowledge base construction system based on a large model provided by one embodiment of the present invention; Figure 3 A schematic diagram of the specific structure of another large-model-based knowledge base construction system provided by one embodiment of the present invention; Figure 4 A schematic structural diagram of a large model-based knowledge base construction device provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0025] The following describes the traditional knowledge base construction scheme based on vector database in related technologies: Corpus vectorization: Use pre-trained models (such as BERT and Sentence-BERT) to convert text into dense vectors and store them in vector databases (such as FAISS and Milvus).

[0026] Search and screening mechanism: Keyword search: Quickly locate documents containing the target entity based on the inverted index (e.g., searching for "spinal surgery" directly returns relevant medical records); Vector similarity retrieval: recall semantically similar entries through vector cosine similarity (e.g., “anesthesia side effects” may recall the case of “postoperative nausea”).

[0027] Rerank optimization: Use the Rarank fine-tuning model to reorder mixed results to improve relevance (for example, when the user enters "anesthesia plan for high-risk patients", the items marked with "ASA grade IV contraindications" are displayed first).

[0028] The following describes the knowledge graph solution in related technologies: The knowledge graph emphasizes explicit relationship modeling and is constructed through the following steps: Entity and relation extraction: Use NER models (such as BIO annotation) to extract entities (such as "bupivacaine" and "neurotoxicity"), and use RE models (such as RENet) to identify relation types (such as "drug-side effect").

[0029] Graph storage and query: Use a graph database (such as Neo4j) to store "entity-relationship-entity" triples; Query example: MATCH (drug)-[:prohibited for]->(disease) WHERE drug = "bupivacaine" RETURN disease.

[0030] Retrieval and reasoning extensions: Infer indirect relationship chains through SPARQL or multi-hop queries (for example, reasoning that "ASA IV patients cannot choose bupivacaine anesthesia" requires joining the path of "ASA IV → contraindicated drugs → bupivacaine").

[0031] Therefore, traditional knowledge base technology has the following defects: Implicit associations are missing: For example, the causal chain "propofol → risk of hypotension → need to monitor blood pressure" needs to be pieced together through multiple searches, and a complete association conclusion cannot be directly output; High cost of dynamic updates: When new knowledge is added, all relevant vectors need to be recalculated and stored, and the Rerank model needs to be retrained to adapt to the new association rules (for example, the launch of a new drug "X" requires manual design of new features); Weak complex query capabilities: For example, when querying "anesthesia plans and contraindications for spinal surgery in patients with renal insufficiency", the system needs to combine keywords ("spinal surgery", "renal insufficiency"), vectors ("anesthesia contraindications") and rerank rules to return results, but may miss embedded related information (for example, the "contraindications" field itself does not include "renal impairment").

[0032] Knowledge graph technology has the following defects: High maintenance cost: Edge relationship management takes up a lot of storage resources (e.g., "each entity pair needs to store a relationship edge separately"), and updates need to ensure full Figure 1 Consistency (for example, deleting "Drug X" requires deleting all associated edges simultaneously).

[0033] Low query efficiency: Complex multi-hop queries (e.g., “find all regimens containing propofol that are contraindicated for patients with renal failure”) rely on traversing the graph, and response time increases exponentially with the depth of the association. Poor dynamic adaptability: Adding new relationship types (such as "long-term side effects of drugs in real-world data") requires manual expansion of the graph structure, which is difficult to meet the needs of rapid iteration scenarios.

[0034] Efficiency bottleneck: Complex associative queries require traversing multiple hops. For example, searching for "anesthetic drugs for patients with normal cardiac function who meet the 'postoperative analgesia plan'" requires multiple hops, resulting in significant delays.

[0035] Therefore, as described in the background art, this application has the dilemma of being highly dependent on complex association structures and being unable to achieve both expressiveness and efficiency.

[0036] Based on this, the embodiments of the present invention provide a knowledge base construction method, system and device based on a large model. Through flattened redundant storage and large-model driven knowledge embedding, it realizes "one-click knowledge unit construction" and "field-level rapid update", eliminating the traditional solution's dependence on complex association structures.

[0037] Figure 1 For a flow chart of a method for constructing a knowledge base based on a large model provided by an embodiment of the present invention, please refer to Figure 1 , this embodiment may include the following steps: Step S101: Acquire original corpus data of a target domain, and annotate the original corpus data with entity types using a term library; the original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents.

[0038] In this embodiment, the target field may be anesthesia specialty, and the original corpus data may include electronic medical records, guidelines, various types of literature, etc. After obtaining the original corpus data, domain terminology enhancement is performed, for example, entity types are annotated through a term library, such as drug names, surgery types, etc.

[0039] Step S102: extract the subject and its preset associated attributes in the original corpus data through the large model.

[0040] The large model can be a general-purpose large language model such as Qwen, DeepSeek, or GPT, specifically, Qwen2.5-72B-Instruct, deepseek-V3, or GPT-4o, as long as it can perform the extraction function. This application does not impose any specific limitations. The large model extracts the subject (e.g., the name of the surgical procedure) and pre-set associated attributes (e.g., medications, contraindications, complications, etc.) from the original corpus data.

[0041] Step S103: Based on any corpus segment, determine the embedded association information construction control instruction of the corpus segment according to the subject and preset association attributes, and input the embedded association information construction control instruction into the large model to obtain candidate knowledge units directly associated with the association attribute field and the subject field.

[0042] After knowledge extraction, control instructions can be input into the large model to obtain candidate knowledge units, wherein the large model can be a neural network model that can execute instructions. This application does not make specific restrictions, and the model can be the same as or different from the first large model.

[0043] For example, for the corpus segment "Spinal surgery requires prohibition of epidural anesthesia for patients with coagulation disorders", the embedded association information construction control instruction can be determined based on the subject (spinal surgery) and preset association attributes (anesthesia plan, contraindications, alternative plans) as "generate a spinal surgery knowledge unit containing anesthesia plan, contraindications, and alternative plans, requiring direct association of fields", thereby enabling the large model to produce candidate knowledge units.

[0044] For example, the control instruction may also be "extract the surgery name and directly related anesthesia plan and contraindications from the following text, and output them in JSON format."

[0045] Step S104: Based on entity type and semantic consistency, perform initial conflict detection on the candidate knowledge units. When conflicts exist among the candidate knowledge units, execute according to the conflict strategy to obtain a knowledge unit set; wherein, the knowledge unit set includes different non-conflicting knowledge units.

[0046] It is determined whether the entity type and semantics in the candidate knowledge unit are consistent. If not, a conflict is determined and step S102 is executed again.

[0047] Step S105: Structuredly store the fields of the knowledge unit in JSON format; the fields include: the main body of current knowledge, directly related knowledge, contraindication list, complication management plan, list of equipment required for surgery, and version control information; vectorize the preset fields of the knowledge unit into the database to obtain a knowledge base.

[0048] For example, deploy the storage system as follows: 1) Structured data storage: json { "entityId": "SUR_OP001", the subject of current knowledge "directKnowledge": [ ... ], directly related knowledge "contraindication": [ ... ], contraindication list "complicationManagement": [ ... ],Complication management plan "equipmentRequirements": [ ... ], a list of equipment required for surgery "versionControl": { ...} version control information } 2) Vectorize fields such as "Subject ID" and "Name" and store them in the database.

[0049] It can be understood that by adopting the technical solution provided in the embodiment of the present invention, the subject and preset associated attributes in the original corpus data are extracted through the big model, so that the candidate knowledge units associated with the subject and the preset associated attributes are obtained through the big model, that is, through flattened redundant storage and big model-driven knowledge embedding, one-click knowledge unit construction is achieved, eliminating the traditional solution's dependence on complex associated structures.

[0050] In some embodiments, the method provided in this embodiment further includes the following steps: The associated attribute fields of the knowledge units of the same subject are expanded. When the updated corpus of any associated attribute field is obtained, the knowledge units related to the updated corpus are updated.

[0051] For example, when updated corpus appears on the same subject (e.g., spinal surgery) (e.g., new ASA classification contraindications), the knowledge units related to the updated corpus are updated.

[0052] For example, if the FDA 2025 guidance updates the contraindications for bupivacaine, include that information in all its associated entities.

[0053] It is understandable that, by adopting the technical solution provided by the embodiment of the present invention, when new corpus appears, only the relevant fields are updated instead of rewriting the entire entry, thereby improving convenience.

[0054] In some embodiments, the method provided in this embodiment further includes the following steps: In response to the assignment instruction, weights are assigned to the preset corpus; When the updated corpus conflicts with the corresponding content in the knowledge unit, the knowledge unit is updated according to the content with a higher weight.

[0055] For example, in this embodiment, all fields are marked with source weights, for example, FDA guideline weight = 9, document weight = 6; when updating corresponding content conflicts in the corpus knowledge unit, the content with the higher weight is updated first.

[0056] For example, if an old entry shows "ASA III operable" and conflicts with the new guidance, the FDA / ASA update should be adopted first, and only the conflicting field should be modified rather than the entire entry being rewritten.

[0057] It is understandable that by adopting the technical solution provided by the embodiment of the present invention, updates can be performed based on high-reliability content.

[0058] In some embodiments, the method provided in this embodiment further includes the following steps: Store the old and new versions of the corpus in parallel and timestamp them; or In response to the field priority setting instruction, the priority of the knowledge unit associated with the field is updated.

[0059] For example, when storing new and old versions of corpus, the new and old versions of the corpus are marked with timestamps. For easy viewing, only one of the same parts of the new and old versions is retained, and for different contents, the new and old corpus and timestamps are retained.

[0060] For example, when the priority of a knowledge item explicitly marked by the user changes (such as marking a specific field as "needs rechecking"), it is dynamically updated.

[0061] It is understandable that retaining the old and new versions and marking them with timestamps can avoid completely erasing old knowledge; when the priority is changed, updates are performed based on the changed priority.

[0062] In some embodiments, the method provided in this embodiment further includes the following steps: When there is duplicate information in any knowledge unit, deduplication is performed.

[0063] For example, the "preoperative fasting requirements" specified in multiple guidelines have been merged into one item.

[0064] In some embodiments, the method provided in this embodiment further includes the following steps: Record all field changes to build a version history table.

[0065] In some embodiments, the method provided in this embodiment further includes the following steps: In response to natural language or fuzzy keywords input by the user, the relevant knowledge units are located through pre-stored entity vector indexes; By screening the structured fields, a complete knowledge unit is determined among the related knowledge units.

[0066] For example, when a user makes a query, they enter a structured query (such as "spinal surgery + ASA level III") or natural speech (such as "select anesthesia plan for patients with renal insufficiency"), and the system quickly locates relevant knowledge units through pre-stored entity vector indexes.

[0067] Within the initial screening results or the entire database, precise searches are performed to directly filter field-level constraints. Responses do not require multi-step splicing, and complete knowledge units that meet all constraints are directly output.

[0068] For example: When a user searches for anesthesia plans for spinal surgery, they can directly and accurately locate the content corresponding to the "anesthesia plan" field in the initial screening result "Spinal Surgery" document. This plan takes spinal surgery as the main body, and builds a document around this subject that contains knowledge such as anesthesia plans, drug contraindications, and complication management. When a user searches for anesthesia plans for spinal surgery, they can directly locate the field content of the anesthesia plan in the document and directly obtain relevant information. Unlike traditional knowledge bases, the information required by users is scattered across different knowledge fragments, and inaccurate answers may be given due to missing context or incomplete content.

[0069] In some embodiments, it further includes: When the number of complete knowledge units is greater than 1, the similarities between different complete knowledge units are determined, and the complete knowledge unit with high credibility is determined as the target knowledge unit and returned to the user; or, When different versions of content exist in a complete knowledge unit, the content with a higher weight is determined as the content of the target knowledge unit and returned to the user.

[0070] For example, when there are multiple versions or similarity results, they are sorted comprehensively by field source weight (such as FDA / authoritative guidelines over literature) and vector similarity, and high-confidence results are returned first.

[0071] In the embodiment of the present invention, the design of the knowledge base storage structure is explained again: Flat redundant storage model: Format example: A "knowledge unit" in JSON format, with related information directly embedded (for example, a surgical plan directly includes contraindicated drug attributes).

[0072] Field design: json { "Subject ID": "SUR_OP001", "Subject Type": "Surgery", "Name": "Spinal Fusion", "Directly related knowledge": { "anesthesia plan": { "Recommended drug combination": ["ropivacaine", "propofol"], "Taboo conditions": [ { "Condition Type": "Patient Status", "Description": "ASA Class IV or higher" }, { "Condition Type": "Laboratory Indicator", "Description": "Coagulation disorder (PT>15 seconds)", "Related drugs": ["epidural anesthetic"] } ], "Alternative options": "If propofol is contraindicated, ketamine + dexmedetomidine should be used instead" }, "Complication Management": { "Management of infection risk": "Preoperative prophylactic use of vancomycin and monitoring of white blood cell count" } }, "Version Control": { "current": "v202504", "lastUpdate": "2025-04-01", "priorVersion": "v202503 Note: Updated FDA 2025 Bupivacaine Contraindications Guidance" } } Storage Policy: Use a relational database (such as MySQL) or NoSQL (such as MongoDB, Elasticsearch) to achieve internal associations through field nesting.

[0073] Fields such as "Subject ID" and "Name" are vectorized to preliminarily screen knowledge units through vector similarity.

[0074] In an embodiment of the present invention, taking the anesthesia specialist knowledge base as an example, this solution can directly merge cross-entity association information such as "anesthesia contraindications for spinal surgery," "complications of anesthetic drugs," and "patient risk assessment" into surgical plan entries, allowing clinicians to obtain complete decision-making basis with a single query, without having to manually integrate multiple scattered pieces of knowledge. This design significantly improves knowledge query efficiency and practical application in scenarios while reducing maintenance complexity, providing a new technical path for the intelligentization of specialized knowledge-intensive fields.

[0075] This embodiment provides a specific embodiment:

[0076] Based on a general inventive concept, the present invention also provides a knowledge base construction system based on a large model, which is used to implement the above method embodiment. Figure 2 A schematic diagram of a knowledge base construction system based on a large model provided by an embodiment of the present invention. Figure 2 As shown, the system provided by the embodiment of the present invention may include: The input layer 21 is used to obtain the original corpus data of the target domain and annotate the entity types of the original corpus data using the term library; the original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents; The processing layer 22 is used to extract the subject and its preset associated attributes from the original corpus data through the large model; based on any corpus segment, determine the embedded associated information construction control instruction of the corpus segment according to the subject and the preset associated attributes, and input the embedded associated information construction control instruction into the large model to obtain candidate knowledge units directly associated with the associated attribute field and the subject field; based on the entity type and semantic consistency, perform initial conflict detection on the candidate knowledge units. When the candidate knowledge units conflict, execute according to the conflict strategy to obtain a knowledge unit set; wherein, the knowledge unit set includes different non-conflicting knowledge units; The storage layer 23 is used to store the fields of the knowledge unit in a structured manner according to the JSON format; the fields include: the main body of the current knowledge, directly related knowledge, a list of contraindications, a complication management plan, a list of equipment required for surgery, and version control information; the preset fields of the knowledge unit are vectorized and stored in the database to obtain a knowledge base.

[0077] In some embodiments, an output layer 24 is further included for locating relevant knowledge units through pre-stored entity vector indexes in response to natural language or fuzzy keywords input by the user; and determining complete knowledge units in the relevant knowledge units by screening structured fields.

[0078] Regarding the system in the above embodiment, the specific manner in which each layer performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0079] For details, see Figure 3 , Figure 3 A specific structural diagram of another large-model-based knowledge base construction system provided for one embodiment of the present invention, wherein the input layer includes target domain corpus (structured + unstructured data); the processing layer includes a preprocessing module, a subject identification and knowledge merging engine; the storage layer may include a flattened redundant storage database; and the output layer includes a knowledge base query management platform.

[0080] The present invention also provides a large model-based knowledge base construction device for implementing the above method embodiment. Figure 4 A schematic diagram of a large model-based knowledge base construction device according to an embodiment of the present invention is shown in FIG. Figure 4 As shown, the large model-based knowledge base construction device of this embodiment includes a processor 31 and a memory 32, and the processor 31 is connected to the memory 32. The processor 31 is used to call and execute the program stored in the memory 32; the memory 32 is used to store the program, and the program is used to execute at least the large model-based knowledge base construction method of the above embodiment.

[0081] The specific implementation scheme of the large-model-based knowledge base construction device provided in the embodiments of this application can refer to the implementation scheme of the large-model-based knowledge base construction method in any of the above embodiments, and will not be repeated here.

[0082] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0083] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0084] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0085] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0086] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0087] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0088] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0089] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0090] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a knowledge base based on a large model, characterized in that: include: Obtaining original corpus data of the target domain and annotating the original corpus data with entity types using a term library; The original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents; Extract the subject and its preset associated attributes from the original corpus data through the large model; Based on any corpus segment, according to the subject and preset association attributes, the embedded association information construction control instruction of the corpus segment is determined, and the embedded association information construction control instruction is input into the large model to obtain candidate knowledge units directly associated with the association attribute field and the subject field; Based on entity type and semantic consistency, the candidate knowledge units are initially detected for conflict. When the candidate knowledge units conflict, the conflict strategy is followed to obtain a knowledge unit set; wherein the knowledge unit set includes different non-conflicting knowledge units; The fields of the knowledge unit are structured and stored in JSON format; the fields include: the main body of current knowledge, directly related knowledge, a list of contraindications, a complication management plan, a list of equipment required for surgery, and version control information; the preset fields of the knowledge unit are vectorized and stored in the database to obtain a knowledge base.

2. The method according to claim 1, characterized in that Also includes: The associated attribute fields of the knowledge units of the same subject are expanded. When the updated corpus of any associated attribute field is obtained, the knowledge units related to the updated corpus are updated.

3. The method according to claim 2, characterized in that Also includes: In response to the assignment instruction, weights are assigned to the preset corpus; When the updated corpus conflicts with the corresponding content in the knowledge unit, the knowledge unit is updated according to the content with a higher weight.

4. The method according to claim 2, characterized in that Also includes: Store the old and new versions of the corpus in parallel and timestamp them. or, In response to the field priority setting instruction, the priority of the knowledge unit associated with the field is updated.

5. The method according to claim 1, characterized in that Also includes: When there is duplicate information in any knowledge unit, deduplication is performed.

6. The method according to claim 1, characterized in that Also includes: Record all field changes to build a version history table.

7. The method according to claim 1, characterized in that Also includes: In response to natural language or fuzzy keywords input by the user, the relevant knowledge units are located through pre-stored entity vector indexes; By screening the structured fields, a complete knowledge unit is determined among the related knowledge units.

8. The method according to claim 7, characterized in that Also includes: When the number of complete knowledge units is greater than 1, the similarity between different complete knowledge units is determined, and the complete knowledge unit with high credibility is determined as the target knowledge unit and returned to the user; or, When different versions of content exist in a complete knowledge unit, the content with a higher weight is determined as the content of the target knowledge unit and returned to the user.

9. A knowledge base construction system based on a large model, characterized in that: include: The input layer is used to obtain the original corpus data of the target domain and annotate the entity types of the original corpus data using the term library; the original corpus data includes different corpora, including electronic medical records, guidelines, and guidance documents; The processing layer is used to extract the subject and its preset associated attributes from the original corpus data through the large model; based on any corpus segment, determine the embedded associated information construction control instruction of the corpus segment according to the subject and preset associated attributes, and input the embedded associated information construction control instruction into the large model to obtain candidate knowledge units directly associated with the associated attribute field and the subject field; based on the entity type and semantic consistency, perform initial conflict detection on the candidate knowledge units, and when there is a conflict between the candidate knowledge units, execute according to the conflict strategy to obtain a knowledge unit set; wherein, the knowledge unit set includes different non-conflicting knowledge units; The storage layer is used to store the fields of the knowledge unit in a structured manner in JSON format; the fields include: the main body of the current knowledge, directly related knowledge, a list of contraindications, a complication management plan, a list of equipment required for surgery, and version control information; the preset fields of the knowledge unit are vectorized and stored in the database to obtain a knowledge base.

10. A knowledge base construction device based on a large model, characterized in that: The device comprises a processor and a memory, wherein the processor is connected to the memory: The processor is configured to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the large model-based knowledge base construction method described in any one of claims 1-8.

Images