Medical information processing method and device, equipment, storage medium and product
By segmenting medical documents and constructing a knowledge graph database, the problem of incomplete retrieval of medical knowledge fragments was solved, resulting in higher recall and answer accuracy.
Patent Information
- Application Number
- CN202510196683.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-21
- Publication Date
- 2026-02-10
AI Technical Summary
When searching for existing medical knowledge fragments, the recall rate is low and incomplete, resulting in insufficient accuracy of the answers.
The original medical documents are segmented to generate data block numbers, content, and summaries, which are then synchronously stored in the database. Based on user queries, the data block content and summaries are retrieved and integrated to construct a knowledge graph database for rewriting queries, thereby improving recall and data integrity.
It improved the recall rate and data integrity of medical knowledge fragments, resulting in more accurate and comprehensive answers.
Smart Images
Figure CN121506433A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical information processing technology, and in particular to a medical information processing method, apparatus, equipment, storage medium and product. Background Technology
[0002] Medical question-answering systems are one of the most popular applications of natural language processing. These systems generate answers based on user input to meet their needs. To improve the accuracy of the answers, the system not only relies on user input but also utilizes retrieved medical knowledge fragments for enhancement, thus providing more accurate and information-rich responses.
[0003] However, current methods for retrieving medical knowledge fragments mainly consider the semantic similarity between user input and the medical knowledge fragment, which may result in incomplete retrieved medical knowledge fragments and low recall rates. Summary of the Invention
[0004] This application provides a medical information processing method, apparatus, equipment, storage medium, and product to address the problem that medical knowledge fragments retrieved by existing technologies may be incomplete and have a low recall rate.
[0005] To achieve the above objectives, embodiments of this application provide a medical information processing method, including:
[0006] The original medical document is divided into blocks to obtain several data blocks;
[0007] The data block is processed into a predefined data structure; wherein the data structure includes at least: data block number, data block content, and a summary of data block content;
[0008] The aforementioned data structures are synchronously stored in several databases;
[0009] Based on user queries, the contents of data blocks and summaries of data block contents in several databases are retrieved to obtain several first target data block contents.
[0010] The contents of several first target data blocks are fused to obtain first medical knowledge information.
[0011] As an improvement to the above scheme, the step of fusing the contents of several first target data blocks to obtain first medical knowledge information includes:
[0012] Identify the contents of several first target data blocks belonging to the same original medical document;
[0013] Sort the contents of several first target data blocks of the same original medical document according to the data block number;
[0014] The contents of several first target data blocks are merged after sorting the contents of the same original medical document.
[0015] As an improvement to the above solution, processing the data block into a predefined data structure includes:
[0016] The data blocks of the same original medical document are numbered according to the order of the document content to obtain the data block number of the data structure.
[0017] Based on the content of the data block, the data block content of the data structure is obtained;
[0018] Using a preset first prompt word, a summary of the data block content of the data structure is generated.
[0019] As an improvement to the above solution, the data structure is generated offline;
[0020] The data structure also includes at least one of the following:
[0021] Document title, document file name, document identifier, document content summary, data block identifier, and data block content quality;
[0022] The data block identifier is obtained by hashing the data block content.
[0023] As an improvement to the above solution, the medical information processing method further includes:
[0024] Based on the original medical documents, a knowledge graph base is constructed;
[0025] The user query is rewritten based on the knowledge graph database to obtain several sub-queries;
[0026] Based on the subquery, the contents of data blocks and summaries of data blocks in several databases are retrieved to obtain several second target data block contents. The contents of several second target data blocks are then fused to obtain second medical knowledge information.
[0027] As an improvement to the above solution, the medical information processing method further includes:
[0028] The knowledge graph database is searched based on the subquery to obtain third-party medical knowledge information.
[0029] To achieve the above objectives, embodiments of this application also provide a medical information processing device, comprising:
[0030] The chunking module is used to divide the original medical document into chunks to obtain several data blocks;
[0031] A processing module is used to process the data block into a predefined data structure; wherein the data structure includes at least: a data block number, a data block content, and a summary of the data block content;
[0032] A synchronization module is used to synchronously store several of the aforementioned data structures into several databases;
[0033] The first retrieval module is used to retrieve data block content and data block content summary of several databases based on user queries, and obtain several first target data block contents.
[0034] The fusion module is used to fuse the contents of several first target data blocks to obtain first medical knowledge information.
[0035] To achieve the above objectives, this application also provides a medical information processing device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the medical information processing method as described above.
[0036] To achieve the above objectives, embodiments of this application also provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the medical information processing method as described above.
[0037] To achieve the above objectives, embodiments of this application also provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the medical information processing method as described above.
[0038] Compared with existing technologies, the medical information processing method, apparatus, device, storage medium, and product provided in this application embodiment divides original medical documents into blocks to obtain several data blocks; processes the data blocks into predefined data structures; wherein, the data structure includes at least: a data block number, data block content, and a summary of data block content; synchronously stores several data structures in several databases; retrieves the data block content and data block content summaries from the several databases based on user queries to obtain several first target data block contents; and fuses the several first target data block contents to obtain first medical knowledge information. Therefore, this application embodiment not only retrieves data block content but also retrieves data block content summaries, which can improve recall and data completeness, ultimately leading to more accurate answers. Attached Figure Description
[0039] Figure 1 This is a flowchart of a medical information processing method provided in an embodiment of this application;
[0040] Figure 2 This is a structural block diagram of a medical information processing device provided in an embodiment of this application;
[0041] Figure 3 This is a structural block diagram of a medical information processing device provided in an embodiment of this application. Detailed Implementation
[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0043] See Figure 1 , Figure 1 This is a flowchart of a medical information processing method provided in an embodiment of this application. The medical information processing method includes:
[0044] S1. Divide the original medical document into blocks to obtain several data blocks;
[0045] S2. Process the data block into a predefined data structure; wherein the data structure includes at least: data block number, data block content, and a summary of data block content;
[0046] S3. Synchronously store several of the data structures into several databases;
[0047] S4. Based on the user query, retrieve the data block content and data block content summary of several databases to obtain several first target data block contents;
[0048] S5. Merge the contents of several first target data blocks to obtain first medical knowledge information.
[0049] This application's embodiments design a data structure for data processing in the medical field, containing multiple fields to support efficient retrieval and analysis:
[0050] Section number: The number assigned to each data chunk. For example, section 1 represents the first data chunk of the original medical document, section 2 represents the second data chunk of the original medical document, and so on. Section 0 indicates no chunking.
[0051] Data block content: The content of each data block chunk, i.e., the content of the segmented parts of the original medical document. When performing searches, the search weight of the data block content summary can be increased to improve recall.
[0052] Data Chunk Summary: This provides a summary of the content of each data chunk, offering a quick overview of its core information and serving as a bridge between different data chunks. In retrieval, increasing the search weight of the data chunk summary can improve recall.
[0053] This embodiment first divides the original medical document into chunks to obtain several data chunks, and then processes the data chunks into predefined data structures. Each data structure includes at least: a data chunk number (section), data chunk content, and a data chunk content summary. Next, these data structures are synchronously stored in several databases. Finally, based on a user query, the data chunk content and data chunk content summary in the several databases are retrieved to obtain several target data chunk contents. These target data chunk contents are then fused to obtain the first medical knowledge information.
[0054] This application does not limit the chunking method; it can be selected according to actual needs, such as through recursive chunking, semantic chunking, and other strategies. Furthermore, by controlling chunk overlap, the recall rate can be improved; chunk_overlap refers to the size of the overlapping area between adjacent data blocks.
[0055] This application embodiment not only retrieves the content of data blocks, but also retrieves the summary of the content of data blocks, which can improve the recall rate and the completeness of the data, and ultimately make the answer more accurate.
[0056] Optionally, the database includes:
[0057] Elasticsearch (ES) library: used for full-text search and fast retrieval.
[0058] Vector Store: Used for similarity search. Vector stores include FAISS, Milvus, and Weaviate. Furthermore, parameters such as nlist and nprobe set in the vector store can also help improve recall; these can be configured according to actual needs.
[0059] In this embodiment, the data structure is synchronously stored in each database, so that each database stores the same data.
[0060] In one optional embodiment, processing the data block into a predefined data structure includes:
[0061] The data blocks of the same original medical document are numbered according to the order of the document content to obtain the data block number of the data structure.
[0062] Based on the content of the data block, the data block content of the data structure is obtained;
[0063] Using a preset first prompt word, a summary of the data block content of the data structure is generated.
[0064] In this embodiment, in offline mode, several data blocks of the same original medical document are numbered according to the order of the document content to obtain data block numbers (sections). In offline mode, a preset first prompt keyword is input into the large model to generate a data block content summary; the data block content is the content of the segmented fragments of the original medical document. This embodiment improves the efficiency and accuracy of retrieval by generating data block numbers and data block content summaries offline.
[0065] For example, the first prompt word is summary_prompt:
[0066] You are an administrator responsible for managing a set of data chunks, which represent groups of sentences on similar topics.
[0067] Please generate a very short, one-sentence summary that tells the reader what the topic of this data block is.
[0068] A good summary will describe the subject of this data block and provide clarifying instructions where necessary.
[0069] Summarize the content where appropriate.
[0070] enter:{}
[0071] Output: .format(chunk_text)
[0072] This allows you to generate a summary (i.e., a summary of the data block content) of all chunks and save it.
[0073] In one alternative embodiment, the data structure is generated offline;
[0074] The data structure also includes at least one of the following:
[0075] Document title, document file name, document identifier, document content summary, data block identifier, and data block content quality;
[0076] The data block identifier is obtained by hashing the data block content.
[0077] It is worth noting that the data structure also includes at least one of the following:
[0078] Document Title: The title of the original medical document, used to identify the document's subject or main content. The title field facilitates quick identification of the document's core information; it can be generated offline and can increase the search weight of the document title during retrieval, improving recall.
[0079] Filename: The filename of the original medical document. It records the name or path of the source file, helping to track and manage the source of the data. Especially when the number of documents is large, the filename facilitates the location and management of the files.
[0080] Document ID: Used to identify which original medical document each data chunk comes from.
[0081] Document content summary outline: This is generated by combining summaries from multiple data blocks. Specifically, in offline mode, preset second prompt words are input into the large model to generate the document content summary outline. During retrieval, the search weight of the document content summary can be increased, enabling article-level global searches and resolving the problem of missing global information.
[0082] Chunk ID: Chunk ID = hash(content). Using the chunk ID as the deduplication criterion, duplicate data blocks are removed during fusion; that is, data blocks with identical content from the same original medical document from different databases are retained, with only one of the duplicates remaining.
[0083] Data block content quality: Used to evaluate the quality level of the data block content. Evaluation criteria can be set according to different needs, such as the accuracy and completeness of information; specific offline annotation and generation are possible.
[0084] For example, the second prompt word outline_prompt:
[0085] You are an editor tasked with creating a clear article outline for a set of general summaries. These summaries summarize the core themes of multiple chunks, each representing a group of sentences revolving around a similar topic. Your task is to construct a comprehensive and logically structured article based on these summaries. Each section of the outline should outline the core content discussed in that section and be arranged in a logical order, ensuring seamless transitions and a smooth overall flow.
[0086] Please follow these steps to generate the outline:
[0087] 1. Integrate and categorize abstracts: Classify abstracts according to similar themes or topics, and group similar content into the same paragraph or section.
[0088] 2. Determine chapter titles: Generate appropriate titles for each section, summarizing the main idea of that section.
[0089] 3. Briefly describe the content of each paragraph: Add one or two sentences to each section to further explain what is being discussed and to lay the groundwork for subsequent arguments.
[0090] 4. Logical Structure: Ensure that the order of each part in the outline is reasonable, forming a logical flow that allows readers to gradually transition from basic information to more complex or in-depth discussions.
[0091] The output format is as follows:
[0092] 1. Chapter Title 1:
[0093] Summary: Briefly explain the main content discussed in this section.
[0094] 2. Chapter Title 2:
[0095] Summary: Briefly explain the main content discussed in this section.
[0096] This prompt can create a hierarchical and logical article outline, and can transform multiple data blocks into a coherent content structure through a summary.
[0097] In one optional embodiment, the database also stores a hierarchical data structure, the hierarchy of which represents the hierarchical structure of original medical documents, data block content summaries, and document content summaries.
[0098] For example, level 0 represents the original content, without any summary or processing; level 1 represents a summary of the original content, typically a summary of data block content; and level 2 represents a summary of summaries, typically a summary of document content. The embodiments of this application can enhance the relevance of the search and improve search accuracy.
[0099] In one optional embodiment, the step of retrieving data block content and data block content summaries from several databases based on user queries to obtain several first target data block contents includes:
[0100] For each of the databases described:
[0101] Based on the user query, the data block content of the database is retrieved, and the retrieved data block content is used as the first target data block content;
[0102] Based on the user query, the database is used to retrieve the summary of data block content, and the data block content corresponding to the retrieved summary of data block content is used as the first target data block content.
[0103] This application embodiment not only retrieves the content of data blocks, but also retrieves the summary of the content of data blocks, which can improve the recall rate and the completeness of the data, and ultimately make the answer more accurate.
[0104] In one optional embodiment, the step of fusing the contents of several first target data blocks to obtain first medical knowledge information includes:
[0105] Identify the contents of several first target data blocks belonging to the same original medical document;
[0106] Sort the contents of several first target data blocks of the same original medical document according to the data block number;
[0107] The contents of several first target data blocks are merged after sorting the contents of the same original medical document.
[0108] It is worth noting that the goal of fusion is to deduplicatively combine target data blocks belonging to the same original medical document from various databases according to the data block number (section), ensuring content integrity and clear source. This embodiment of the application achieves article-level fusion and source tracing, improving the accuracy of retrieval and the reliability of source tracing.
[0109] In a specific example, a search is performed in a vector database and an Elasticsearch library:
[0110] Vector database retrieval: Using Milvus to retrieve summary and content:
[0111] Chunked Documents: Records the searched data blocks by title and section, with the data block source marked as "vector".
[0112] Unblocked documents: If a document is not blocked, its content is saved directly, and the data block source is marked as "vector".
[0113] ES library retrieval: Using a match query to search for summary and content using ik_smart analyzer:
[0114] Chunked Documents: Records the searched data blocks by title and section, with the data block source marked as "ES".
[0115] Unblocked documents: If a document is not blocked, its content is saved directly, and the data block source is marked as "ES".
[0116] Content integration:
[0117] Unblocked documents: Unblocked documents are processed first. If a document exists in the search results of both the vector database and the ES database, it is retained only once and marked as originating from "ES / vector".
[0118] Chunked document: Content is concatenated sequentially according to the section order to ensure continuity.
[0119] For example:
[0120] Original medical document A: This document appears in the search results of both the vector database and the ES database, and is not segmented. The final result is to retain the document only once, with the data block source marked as "ES / vector".
[0121] Original medical document B: Sections 1 and 3 were retrieved from the vector database. Sections 2 and 4 were retrieved from the ES database. Therefore, the corresponding data blocks are concatenated according to the order of sections 1, 2, 3, and 4.
[0122] In an optional embodiment, the medical information processing method further includes:
[0123] Based on the original medical documents, a knowledge graph base is constructed;
[0124] The user query is rewritten based on the knowledge graph database to obtain several sub-queries;
[0125] Based on the subquery, the contents of data blocks and summaries of data blocks in several databases are retrieved to obtain several second target data block contents. The contents of several second target data blocks are then fused to obtain second medical knowledge information.
[0126] This application utilizes graphrag technology to extract medical entities from original medical documents and constructs a knowledge graph library, which stores medical entities and relationships. This application leverages the knowledge graph library to rewrite the user's natural language query, decomposing the original query into multi-step operations, i.e., several-word queries, thereby improving recall. Furthermore, using both first and second medical knowledge information as reasoning basis enhances the system's reasoning ability, resulting in more accurate and comprehensive generated answers.
[0127] Specifically, the Aho-Corasick state machine string matching algorithm is used to extract multiple keywords from the user query. This algorithm constructs a finite state automaton (FSA), which allows matching multiple patterns or keywords simultaneously in a single text scan. The extracted keywords are then used to extract descriptions and relationships between them from a knowledge graph database, generating several subqueries.
[0128] For example, the question "How to treat nephropathy, a complication of type 2 diabetes" could be rewritten as "What medications treat type 2 diabetes" or "Which medications are suitable for treating nephropathy caused by type 2 diabetes".
[0129] In an optional embodiment, the medical information processing method further includes:
[0130] The knowledge graph database is searched based on the subquery to obtain third-party medical knowledge information.
[0131] This application embodiment constructs a knowledge graph database, connecting various entities (such as diseases, drugs, and treatment methods) together. Reasoning is performed on the graph structure, effectively addressing problems requiring multiple reasoning steps, resolving multi-hop recall issues, and improving retrieval efficiency. Furthermore, using both first and third medical knowledge information as the basis for reasoning enhances the system's reasoning capabilities, resulting in more accurate and comprehensive generated answers.
[0132] For example: Enter the user query: How to treat kidney disease, a complication of type 2 diabetes.
[0133] Subquery 1: What medications are available for treating type 2 diabetes?
[0134] Recall Result 1: Metformin, sulfonylureas, GLP-1 receptor agonists, etc.
[0135] Subquery 2: What medications are available for treating diabetic nephropathy?
[0136] Recall Result 2: ACE inhibitors and ARB drugs.
[0137] Final multi-beat results: It is recommended to use metformin (for type 2 diabetes) in combination with ACE inhibitors or ARBs (for kidney disease).
[0138] The example above uses the condition (type 2 diabetes) to find its related complications (nephropathy), and then finds drugs suitable for treating these complications, thus solving the multi-hop recall problem and improving search efficiency.
[0139] This application provides a medical information processing method that divides original medical documents into blocks to obtain several data blocks; processes these data blocks into predefined data structures; wherein each data structure includes at least: a data block number, data block content, and a summary of the data block content; synchronously stores these data structures in several databases; retrieves the data block content and the summary of the data block content in the several databases based on user queries to obtain several first target data block contents; and fuses the contents of the several first target data blocks to obtain first medical knowledge information. Therefore, this application not only retrieves the data block content but also the summary of the data block content, which can improve recall and data completeness, ultimately leading to more accurate answers.
[0140] See Figure 2 , Figure 2 This is a structural block diagram of a medical information processing device 10 provided in an embodiment of this application. The medical information processing device 10 includes:
[0141] Blocking module 11 is used to divide the original medical document into blocks to obtain several data blocks;
[0142] Processing module 12 is used to process the data block into a predefined data structure; wherein the data structure includes at least: data block number, data block content, and a summary of data block content;
[0143] Synchronization module 13 is used to synchronously store several of the data structures into several databases;
[0144] The first retrieval module 14 is used to retrieve the data block content and data block content summary of several databases based on user queries, and obtain several first target data block contents.
[0145] The fusion module 15 is used to fuse the contents of several first target data blocks to obtain first medical knowledge information.
[0146] Optionally, the fusion module 15 is specifically used for:
[0147] Identify the contents of several first target data blocks belonging to the same original medical document;
[0148] Sort the contents of several first target data blocks of the same original medical document according to the data block number;
[0149] The contents of several first target data blocks are merged after sorting the contents of the same original medical document.
[0150] Optionally, the processing module 12 is specifically used for:
[0151] The data blocks of the same original medical document are numbered according to the order of the document content to obtain the data block number of the data structure.
[0152] Based on the content of the data block, the data block content of the data structure is obtained;
[0153] Using a preset first prompt word, a summary of the data block content of the data structure is generated.
[0154] Optionally, the data structure is generated offline;
[0155] The data structure also includes at least one of the following:
[0156] Document title, document file name, document identifier, document content summary, data block identifier, and data block content quality;
[0157] The data block identifier is obtained by hashing the data block content.
[0158] Optionally, the medical information processing device 10 further includes:
[0159] The second retrieval module is used to construct a knowledge graph base based on the original medical documents; rewrite the user query based on the knowledge graph base to obtain several sub-queries; retrieve the data block content and data block content summary of several databases based on the sub-queries to obtain several second target data block contents; and fuse the several second target data block contents to obtain second medical knowledge information.
[0160] Optionally, the medical information processing method apparatus 10 further includes:
[0161] The third retrieval module retrieves third medical knowledge information from the knowledge graph database based on the subquery.
[0162] It is worth noting that the working process of each module in the medical information processing device 10 described in this application embodiment can refer to the working process of the medical information processing method described in the above embodiment, and will not be repeated here.
[0163] This application provides a medical information processing device 10 that divides original medical documents into blocks to obtain several data blocks; processes the data blocks into predefined data structures; wherein the data structures at least include: a data block number, data block content, and a summary of data block content; synchronously stores several data structures in several databases; retrieves the data block content and data block content summaries from the several databases based on user queries to obtain several first target data block contents; and fuses the several first target data block contents to obtain first medical knowledge information. Therefore, this application embodiment not only retrieves data block content but also retrieves data block content summaries, which can improve recall and data completeness, ultimately leading to more accurate answers.
[0164] Furthermore, this application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the medical information processing method as described in any of the above embodiments.
[0165] Furthermore, this application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the medical information processing method as described in any of the above embodiments.
[0166] See Figure 3 , Figure 3 This is a structural block diagram of a medical information processing device 20 provided in an embodiment of this application. The medical information processing device 20 includes: a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the steps in the above-described medical information processing method embodiments. Alternatively, when the processor 21 executes the computer program, it implements the functions of each module / unit in the above-described device embodiments.
[0167] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the medical information processing device 20.
[0168] The medical information processing device 20 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the medical information processing device 20 and does not constitute a limitation on the device. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the medical information processing device 20 may also include input / output devices, network access devices, buses, etc.
[0169] The processor 21 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the medical information processing device 20, connecting all parts of the medical information processing device 20 via various interfaces and lines.
[0170] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the medical information processing device 20 by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0171] If the modules / units integrated in the medical information processing device 20 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 21, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.
[0172] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.
[0173] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.
Claims
1. A medical information processing method, characterized in that, include: The original medical document is divided into blocks to obtain several data blocks; The data block is processed into a predefined data structure; wherein the data structure includes at least: data block number, data block content, and a summary of data block content; The aforementioned data structures are synchronously stored in several databases; Based on user queries, the contents of data blocks and summaries of data block contents in several databases are retrieved to obtain several first target data block contents. The contents of several first target data blocks are fused to obtain first medical knowledge information.
2. The medical information processing method as described in claim 1, characterized in that, The process of fusing the contents of several first target data blocks to obtain first medical knowledge information includes: Identify the contents of several first target data blocks belonging to the same original medical document; Sort the contents of several first target data blocks of the same original medical document according to the data block number; The contents of several first target data blocks are merged after sorting the contents of the same original medical document.
3. The medical information processing method as described in claim 1, characterized in that, The step of processing the data block into a predefined data structure includes: The data blocks of the same original medical document are numbered according to the order of the document content to obtain the data block number of the data structure. Based on the content of the data block, the data block content of the data structure is obtained; Using a preset first prompt word, a summary of the data block content of the data structure is generated.
4. The medical information processing method as described in claim 1, characterized in that, The data structure was generated offline. The data structure also includes at least one of the following: Document title, document file name, document identifier, document content summary, data block identifier, and data block content quality; The data block identifier is obtained by hashing the data block content.
5. The medical information processing method as described in claim 1, characterized in that, The medical information processing method further includes: Based on the original medical documents, a knowledge graph base is constructed; The user query is rewritten based on the knowledge graph database to obtain several sub-queries; Based on the subquery, the contents of data blocks and summaries of data blocks in several databases are retrieved to obtain several second target data block contents. The contents of several second target data blocks are then fused to obtain second medical knowledge information.
6. The medical information processing method as described in claim 5, characterized in that, The medical information processing method further includes: The knowledge graph database is searched based on the subquery to obtain third-party medical knowledge information.
7. A medical information processing device, characterized in that, include: The chunking module is used to divide the original medical document into chunks to obtain several data blocks; A processing module is used to process the data block into a predefined data structure; wherein the data structure includes at least: a data block number, a data block content, and a summary of the data block content; A synchronization module is used to synchronously store several of the aforementioned data structures into several databases; The first retrieval module is used to retrieve data block content and data block content summary of several databases based on user queries, and obtain several first target data block contents. The fusion module is used to fuse the contents of several first target data blocks to obtain first medical knowledge information.
8. A medical information processing device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the medical information processing method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the medical information processing method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, It includes a computer program / instruction that, when executed by a processor, implements the medical information processing method as described in any one of claims 1 to 6.