Systematized literature review method and system based on large language model

Abstract

The invention provides a systematized literature review method and system based on a large language model. The systematized literature review method based on the large language model comprises the following steps that an SLR scheme first draft is generated according to project requirements based on the large language model; confirming the first draft of the SLR scheme to obtain an SLR research scheme; performing literature retrieval and analysis based on the SLR research scheme to obtain a preliminary retrieval result; and modifying the SLR research scheme based on the preliminary retrieval result to obtain a literature review result. According to the technical scheme, the efficiency and reliability of the SLR are effectively improved, and important support is provided for the scientific research and medical field.

Description

Background Art

[0002] Systematic Literature Review (SLR) is a scientific method that uses rigorous methodology to retrieve, screen and analyze relevant literature to answer specific research questions. The traditional SLR process is time-consuming and labor-intensive, and is prone to inconsistent results due to human factors. Summary of the invention

[0003] The present application provides a systematic literature review method and system based on a large language model to improve the efficiency, accuracy and repeatability of the systematic literature review.

[0004] In the first aspect, a systematic literature review method based on a large language model is provided, comprising the following steps:

[0005] Generate a first draft of the SLR solution based on the large language model according to project requirements;

[0006] Confirm the draft of the SLR plan to obtain the SLR research plan;

[0007] Based on the SLR research plan, literature search and analysis were performed to obtain preliminary search results;

[0008] Based on the preliminary search results, the SLR research plan was modified to obtain the literature review results.

[0009] In the above technical scheme, a preliminary draft of the SLR plan is generated according to the project requirements based on a large language model; the preliminary draft of the SLR plan is confirmed to obtain an SLR research plan; literature search and analysis are performed based on the SLR research plan to obtain preliminary search results; the SLR research plan is modified based on the preliminary search results to obtain literature review results; the efficiency and reliability of the SLR are effectively improved, providing important support for scientific research and medical fields.

[0010] In a specific implementation plan, the steps of generating a first draft of the SLR solution based on the large language model according to the project requirements are as follows:

[0011] The SLR Protocol based on the large language model generates the relevant SLR plan draft according to the project requirements.

[0012] In a specific implementation scheme, the steps of generating the relevant SLR scheme draft based on the SLR Protocol of the large language model according to the project requirements specifically include:

[0013] Generate background and research rationale based on large language models;

[0014] Generate specific goals based on the large language model;

[0015] Generative methods based on large language models;

[0016] Generate retrieval strategies based on large language models.

[0017] In a specific implementation scheme, the steps of generating background and research reasons based on the large language model specifically include:

[0018] Enter the title of your study;

[0019] According to the research title, analysis and context generation are performed based on a large language model;

[0020] Identify and reason about research questions based on large language models;

[0021] Provide complete background and rationale for your research.

[0022] In a specific implementation scheme, the step of generating a specific target based on a large language model specifically includes:

[0023] Through a large model based on the Transformer architecture, relevant information in the field is extracted according to the research topic or keywords and a description of the purpose of the review is generated;

[0024] Using a pre-trained language model, the research elements are analyzed based on the input research background information and the PICO / PEO framework, and a standardized expression of the main research questions is generated through semantic reasoning;

[0025] Through the named entity recognition technology of the large model, the relevant features of the research population are extracted from the domain corpus, and the definition of the research population is generated by combining the contextual semantic relationship;

[0026] Use deep learning-based language generation technology to parse intervention program keywords and retrieve relevant information from the database to generate standardized descriptions of intervention measures;

[0027] Through semantic matching technology, the research requirements input by users are parsed based on a large language model, and comparison group options are extracted and definitions are generated;

[0028] Through the indicator recommendation module based on the pre-trained large model, the input research background and objectives are analyzed, the main outcome indicators are extracted, and customized outcome indicator descriptions are generated based on user needs.

[0029] In a specific implementation scheme, the steps of the large language model generation method specifically include:

[0030] Using a large language model, we generate inclusion and exclusion criteria by semantically parsing the domain corpus and the research objectives input by users;

[0031] The large model was used to analyze the research topics and combine the research design types to generate the included research design criteria and exclusion types;

[0032] Through the large language model, combined with the medical knowledge base and semantic analysis technology, the research background is analyzed and the detailed characteristics of the research object are generated;

[0033] Use large language models to extract interventions or exposures and generate structured descriptions of doses, dosing schedules, and other parameters.

[0034] The domain literature and experimental design guidelines are parsed through a large language model to generate comparison group definitions;

[0035] Based on a large language model and domain knowledge base, the primary and secondary outcome indicators related to the research topic are automatically extracted to generate structured content;

[0036] The research needs and literature metadata are parsed through a large model to generate screening criteria for language and publication date.

[0037] In a specific implementation scheme, the steps of generating a search strategy based on a large language model are as follows:

[0038] The literature search strategy generation technology based on the large language model is used to generate a complete search strategy.

[0039] In a specific implementation scheme, the complete search strategy covers database recommendation, keyword optimization, grey literature expansion and restriction condition definition.

[0040] In a specific embodiment, the format of the SLR study protocol includes the PRISMA format and the COCHRANE format.

[0041] Secondly, a systematic literature review system based on a large language model is provided, including:

[0042] SLR draft module, used to generate SLR draft according to project requirements based on the large language model;

[0043] An SLR research plan module is used to confirm the SLR plan draft and obtain the SLR research plan;

[0044] A literature search and analysis module is used to search and analyze literature based on the SLR research plan to obtain preliminary search results;

[0045] A scheme modification module is used to modify the SLR research scheme based on the preliminary search results to obtain literature review results.

[0046] In the above technical scheme, a preliminary draft of the SLR plan is generated according to the project requirements based on a large language model; the preliminary draft of the SLR plan is confirmed to obtain an SLR research plan; literature search and analysis are performed based on the SLR research plan to obtain preliminary search results; the SLR research plan is modified based on the preliminary search results to obtain literature review results; the efficiency and reliability of the SLR are effectively improved, providing important support for scientific research and medical fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flowchart of a systematic literature review method based on a large language model provided in an embodiment of the present application;

[0048] Figure 2 Another block diagram of the structure of a systematic literature review system based on a large language model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The present application is further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present application will become clearer and more specific.

[0050] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0051] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0052] To facilitate understanding of the systematic literature review method and system based on a large language model provided in the embodiment of the present application, its application scenario is first explained. The systematic literature review method and system based on a large language model provided in the embodiment of the present application are used to improve the efficiency, accuracy and repeatability of the systematic literature review. Systematic literature review is a scientific method for answering specific research questions by searching, screening and analyzing relevant literature through strict methodology. The traditional SLR process is time-consuming and labor-intensive, and is prone to inconsistent results due to human factors. For this reason, the embodiment of the present application provides a systematic literature review method and system based on a large language model to improve the efficiency, accuracy and repeatability of the systematic literature review. The following is a detailed description of the embodiments in conjunction with the specific drawings.

[0053] refer to Figure 1 and Figure 2 , Figure 1 A flowchart of a systematic literature review method based on a large language model provided in an embodiment of the present application; Figure 2Another block diagram of the structure of a systematic literature review system based on a large language model provided in an embodiment of the present application.

[0054] exist Figure 1 In the embodiment of the present application, a systematic literature review method based on a large language model is provided, comprising the following steps:

[0055] Generate a first draft of the SLR solution based on the large language model according to project requirements;

[0056] Confirm the draft of the SLR plan to obtain the SLR research plan;

[0057] Based on the SLR research plan, literature search and analysis were performed to obtain preliminary search results;

[0058] Based on the preliminary search results, the SLR research plan was modified to obtain the literature review results.

[0059] In the above technical scheme, a preliminary draft of the SLR plan is generated according to the project requirements based on a large language model; the preliminary draft of the SLR plan is confirmed to obtain an SLR research plan; literature search and analysis are performed based on the SLR research plan to obtain preliminary search results; the SLR research plan is modified based on the preliminary search results to obtain literature review results; the efficiency and reliability of the SLR are effectively improved, providing important support for scientific research and medical fields.

[0060] In a specific implementation plan, the steps of generating a first draft of the SLR solution based on the large language model according to the project requirements are as follows:

[0061] The SLR Protocol based on the large language model generates the relevant SLR plan draft according to the project requirements.

[0062] In a specific implementation scheme, the steps of generating the relevant SLR scheme draft based on the SLR Protocol of the large language model according to the project requirements specifically include:

[0063] Generate background and research rationale based on large language models;

[0064] Generate specific goals based on the large language model;

[0065] Generative methods based on large language models;

[0066] Generate retrieval strategies based on large language models.

[0067] In a specific implementation scheme, the steps of generating background and research reasons based on the large language model specifically include:

[0068] Enter the title of your study;

[0069] According to the research title, analysis and context generation are performed based on a large language model;

[0070] Identify and reason about research questions based on large language models;

[0071] Provide complete background and rationale for your research.

[0072] In a specific implementation scheme, the step of generating a specific target based on a large language model specifically includes:

[0073] Through a large model based on the Transformer architecture, relevant information in the field is extracted according to the research topic or keywords and a description of the purpose of the review is generated;

[0074] Using a pre-trained language model, the research elements are analyzed based on the input research background information and the PICO / PEO framework, and a standardized expression of the main research questions is generated through semantic reasoning;

[0075] Through the named entity recognition technology of the large model, the relevant features of the research population are extracted from the domain corpus, and the definition of the research population is generated by combining the contextual semantic relationship;

[0076] Use deep learning-based language generation technology to parse intervention program keywords and retrieve relevant information from the database to generate standardized descriptions of intervention measures;

[0077] Through semantic matching technology, the research requirements input by users are parsed based on a large language model, and comparison group options are extracted and definitions are generated;

[0078] Through the indicator recommendation module based on the pre-trained large model, the input research background and objectives are analyzed, the main outcome indicators are extracted, and customized outcome indicator descriptions are generated based on user needs.

[0079] In a specific implementation scheme, the steps of the large language model generation method specifically include:

[0080] Using a large language model, we generate inclusion and exclusion criteria by semantically parsing the domain corpus and the research objectives input by users;

[0081] The large model was used to analyze the research topics and combine the research design types to generate the included research design criteria and exclusion types;

[0082] Through the large language model, combined with the medical knowledge base and semantic analysis technology, the research background is analyzed and the detailed characteristics of the research object are generated;

[0083] Use large language models to extract interventions or exposures and generate structured descriptions of doses, dosing schedules, and other parameters.

[0084] The domain literature and experimental design guidelines are parsed through a large language model to generate comparison group definitions;

[0085] Based on a large language model and domain knowledge base, the primary and secondary outcome indicators related to the research topic are automatically extracted to generate structured content;

[0086] The research needs and literature metadata are parsed through a large model to generate screening criteria for language and publication date.

[0087] In a specific implementation scheme, the steps of generating a search strategy based on a large language model are as follows:

[0088] The literature search strategy generation technology based on the large language model is used to generate a complete search strategy.

[0089] In a specific implementation scheme, the complete search strategy covers database recommendation, keyword optimization, grey literature expansion and restriction condition definition.

[0090] In a specific embodiment, the format of the SLR study protocol includes the PRISMA format and the COCHRANE format.

[0091] refer to Figure 1 Specifically, the systematic literature review method based on the large language model includes:

[0092] 1. Workflow

[0093] The present application provides an optimized SLR workflow based on a large language model, which specifically includes the following steps:

[0094] 1.1 Form a preliminary draft of the SLR plan according to project requirements:

[0095] Generate a first draft of the SLR plan based on the research objectives, research questions, search keywords, and literature query standards provided by the researchers, clarify the search keywords and standards, and generate the relevant first draft of the SLR plan based on the SLR Protocol based on the large language model technology.

[0096] 1.2 Discuss and confirm the SLR plan with the investigator:

[0097] According to the project requirements and the information in the first point above, an executable SLR research plan is formed. The research plan can consider using the two internationally recognized formats of PRISMA (https: / / www.prisma-statement.org / ) or COCHRANE (https: / / www.cochrane.org / ).

[0098] 1.3 Literature search and analysis according to the protocol:

[0099] 1.3.1 Carry out literature search according to the determined SLR scheme and make a preliminary analysis of the search results. During the work process, the search results of each step should be recorded (LOG) to facilitate error detection during the work and as evidence of the quality of the subsequent work. This information and the SLR scheme can ensure the accuracy of the literature search work and the repeatability of its results.

[0100] 1.3.2 Generate a flow diagram of the literature search including the reasons for exclusion. A PRISMA flow diagram can be used as a reference. A flow diagram shows the flow of information during the different stages of a systematic review. It indicates the number of records identified, the number of included and excluded, and the reasons for exclusion. Different templates can be used depending on the type of review (new or updated) and the sources used to identify the studies: see the following link: https: / / www.prisma-statement.org / prisma-2020-flow-diagram.

[0101] 1.4 Discuss the preliminary search results with the researchers and modify the plan:

[0102] Discuss the search results with the researchers and adjust the relevant contents of the plan, such as keywords and search scope, based on the discussion results. After the modification is complete, proceed to the subsequent workflow (extract key information from the confirmed literature search results for quantitative processing and other subsequent steps).

[0103] 2. Key technologies in the work process

[0104] 2.1 SLR Protocol Generation Based on Large Language Model Technology

[0105] The input checklist of the systematic literature review (SLR) protocol requires researchers to input the following specific content and the innovations based on the large language model:

[0106] 2.1.1 Title

[0107] Example:

[0108] Efficacy and safety of Mavacamten in the treatment of obstructive hypertrophic cardiomyopathy: a systematic review of randomized controlled trials.

[0109] 2.1.2 Background and research reasons

[0110] 2.1.2.1 Background

[0111] An overview of the research topic and the need for conducting a review.

[0112] 2.1.2.2 Research reasons

[0113] State the problem, knowledge gap, or clinical question and make clear what the review will address.

[0114] The process of generating specific “background and research reasons” based on big model technology:

[0115] The specific steps include:

[0116] 1) Input the research title;

[0117] 2) Large model analysis and background generation: The system automatically extracts background information related to the research topic from massive literature, such as disease characteristics, patient population distribution, and deficiencies in existing treatment strategies, through a knowledge extraction module based on a pre-trained large model. 3) Research problem identification and reasoning: The system uses large model prompt word generation technology to predict possible research gaps and key issues based on the input preliminary research direction using a neural network, and generates an accurate description of the research reasons.

[0118] 4) Automatically optimize generated content and output complete background and research reasons.

[0119] The system uses a large model based on the Transformer architecture to integrate multi-source data across domains (such as literature statistics and clinical trial data), and combines it with contextual semantic reasoning to automatically generate logical background descriptions.

[0120] 2.1.3 Specific objectives

[0121] 2.1.3.1 Review Purpose

[0122] Clarify the purpose of the review.

[0123] Example: To evaluate the efficacy and safety of Mavacamten in patients with obstructive hypertrophic cardiomyopathy (oHCM).

[0124] 2.1.3.2 Research Questions

[0125] Clarify the main research questions through a framework (such as PICO / PEO).

[0126] Example: What are the efficacy and safety of Mavacamten compared with placebo or standard care in adults with obstructive hypertrophic cardiomyopathy? 2.1.3.3

[0128] Study population

[0129] Example: Patients aged ≥18 years with obstructive hypertrophic cardiomyopathy (oHCM) 2.1.3.4

[0131] Intervention (or exposure)

[0132] Example: Mavacamten for oHCM without dose or duration restrictions. 2.1.3.5

[0134] Comparison group (if applicable)

[0135] Examples: Placebo or conventional treatment, such as beta-blockers, calcium channel blockers, or disopyramide. 2.1.3.6

[0137] Outcome Measures

[0138] Examples: including primary indicators, other efficacy indicators and safety indicators, etc.

[0139] The process of generating “specific goals” based on big model technology:

[0140] Specifically including the following steps:

[0141] 1) The system uses a large model based on the Transformer architecture to extract field-related information and generate a description of the review purpose based on the research topic or keywords (such as "Mavacamten" and "oHCM") entered by the user, ensuring the clarity and academic nature of the goal statement.

[0142] 2) This system uses a pre-trained language model to parse research elements based on the input research background information and the PICO / PEO framework, and generates standardized expressions of major research questions through semantic reasoning to support research design and retrospective analysis.

[0143] 3) The system uses the named entity recognition (NER) technology of a large model to extract relevant features of the research population from the domain corpus, and combines the contextual semantic relationships to generate a definition of the research population that conforms to domain specifications.

[0144] 4) This system uses deep learning-based language generation technology to parse the intervention plan keywords (such as "Mavacamten") entered by the user, retrieve relevant information from the database, and generate a standardized description of the intervention measures.

[0145] 5) The system uses semantic matching technology and a large language model to parse the research requirements input by the user, extract possible comparison group options from standard treatment guidelines and clinical trial literature, and generate definitions.

[0146] 6) The system uses an indicator recommendation module based on a pre-trained large model to parse the input research background and objectives, extract key outcome indicators (such as "changes in Valsalva LVOT gradient at rest and after exercise") from field literature, and generate customized outcome indicator descriptions based on user needs.

[0147] 2.1.4 Methods

[0148] 2.1.4.1 Inclusion and Exclusion Criteria

[0149] 2.1.4.2 Study design / type

[0150] Define criteria for inclusion and exclusion of study types (e.g. RCTs, cohort studies, qualitative studies).

[0151] 2.1.4.3 Study subjects / population

[0152] Characteristics of the population (e.g., age, disease status, study setting).

[0153] 2.1.4.4 Interventions / Exposures

[0154] Describe the specific details of the intervention, treatment, or exposure being studied.

[0155] 2.1.4.5 Comparison Group

[0156] The types of comparison groups allowed in the study (if applicable).

[0157] 2.1.4.6 Outcome indicators

[0158] Specific outcomes to be measured (e.g. mortality, quality of life, efficacy).

[0159] 2.1.4.7 Language / Release Date

[0160] Language and publication date restrictions.

[0161] Example: Only studies published in English with a publication date range of 2014 to 2024 were included.

[0162] The process of generating specific "methods" based on large model technology:

[0163] The specific steps include:

[0164] 1) The system uses a large language model to perform semantic analysis on the domain corpus and the research objectives input by the user, and automatically generates inclusion and exclusion criteria, including specific details such as research design, research subject characteristics and intervention measures.

[0165] 2) This system uses a large model to analyze the research topics input by users, and combines the research design types in the literature to automatically generate included research design criteria (such as RCT) and exclusion types to ensure the scientificity and standardization of the research methods.

[0166] 3) The system uses a large language model, combined with a medical knowledge base and semantic analysis technology, to parse the research background input by the user and generate detailed characteristics of the research object, including age range, disease diagnosis criteria and research environment description, to ensure the accuracy and standardization of the definition of the research population.

[0167] 4) This system uses a large language model to extract interventions or exposure content related to the input topic from the domain literature and generate structured descriptions covering dosage, administration time and other parameters to meet the needs of different study designs.

[0168] 5) The system uses a large model to parse field literature and trial design guidelines to generate applicable comparison group definitions, covering placebo or conventional treatment options, to support users' applications in diverse research scenarios.

[0169] 6) Based on a large language model and domain knowledge base, this system automatically extracts primary and secondary outcome indicators related to the research topic and generates structured content, including specific indicators such as efficacy, safety, and patient-reported outcomes, to ensure the comprehensiveness and accuracy of the outcome definition.

[0170] 7) The system parses research needs and literature metadata through a large model, automatically generates language and publication date screening criteria (such as "only include English studies"), and supports users to dynamically adjust the criteria.

[0171] 2.1.5 Search strategy

[0172] 2.1.5.1 Database to be searched

[0173] List electronic databases (e.g. PubMed, Cochrane Library, Embase).

[0174] 2.1.5.2 Search terms / keywords

[0175] The search string or term to use, including Boolean operators.

[0176] Example: ("oHCM"OR"HCM"OR"HOCM"OR"Obstructive Hypertrophiccardiomyopathy"OR"Hypertrophic cardiomyopathy"OR"hypertrophic obstructivecardiomyopathy")AND("Mavacamten"OR"cardiac myosin inhibitor"OR"Camzyos"OR"myosin inhibitor"OR"MYK-461")

[0177] 2.1.5.3 Other search methods

[0178] Manual searches, grey literature, reference lists, trial registers, etc.

[0179] 2.1.5.4 Application Restrictions

[0180] Restrictions on language, publication date, study type, etc.

[0181] 2.1.5.5 Searching for Document Records

[0182] How to log search results, including database and platform details.

[0183] Example: Search results, including details of database and platform, will be documented and tracked using a PRISMA flow diagram.

[0184] The process of generating a specific "retrieval strategy" based on large model technology:

[0185] This system uses literature search strategy generation technology based on a large language model to automatically generate a complete search strategy covering database recommendations, keyword optimization, gray literature expansion, and restriction definition from user input prompts. Each module is implemented through semantic parsing, named entity recognition, and knowledge graph technology. The generated content is highly complete, accurate, and traceable, significantly improving the efficiency and quality of systematic literature review.

[0186] 2.1.6 Other optional input checklists for the SLR protocol include:

[0187] The following contents should be filled in based on the large model or manually supplemented, including data management, study selection process, data extraction, bias risk (quality assessment), data synthesis and analysis, reporting bias assessment, ethical considerations, dissemination plan, timetable and budget, references, and others.

[0188] In a specific implementation scheme, the systematic literature review method based on a large language model comprises the following steps:

[0189] Generate a first draft of the SLR plan:

[0190] The system analyzes the input research objectives and calls the large language model to automatically generate SLR solutions, including background descriptions, research questions, retrieval strategies, and other content.

[0191] Researchers cross-confirmed:

[0192] The protocol was modified in real time with researchers through an online interactive platform and finally confirmed in a standardized format.

[0193] Literature search and screening:

[0194] The system connects to multiple databases (such as PubMed and Embase), obtains literature data based on optimized search strategies, and records the search steps.

[0195] Quality Control and Data Extraction:

[0196] The system generates the final document set based on the recorded logs and customer feedback. Note that the logs are recorded for each step of the search results as a basis for the quality of work and the repeatability of the results.

[0197] Results submission and presentation:

[0198] The output includes the PRISMA flow chart (a visual process for literature screening and troubleshooting is automatically generated according to the PRISMA flow chart standards), literature collection, and preliminary conclusions.

[0199] In this embodiment, a systematic literature review workflow based on large language model optimization is provided, covering key links such as draft generation, protocol confirmation, literature retrieval, and achievement acquisition. This method can effectively improve the efficiency and reliability of SLR and provide important support for scientific research and medical fields.

[0200] exist Figure 2 In the embodiment of the present application, a systematic literature review system based on a large language model is provided, including:

[0201] SLR draft module, used to generate SLR draft according to project requirements based on the large language model;

[0202] An SLR research plan module is used to confirm the SLR plan draft and obtain the SLR research plan;

[0203] A literature search and analysis module is used to search and analyze literature based on the SLR research plan to obtain preliminary search results;

[0204] A scheme modification module is used to modify the SLR research scheme based on the preliminary search results to obtain literature review results.

[0205] In the above technical scheme, a preliminary draft of the SLR plan is generated according to the project requirements based on a large language model; the preliminary draft of the SLR plan is confirmed to obtain an SLR research plan; literature search and analysis are performed based on the SLR research plan to obtain preliminary search results; the SLR research plan is modified based on the preliminary search results to obtain literature review results; the efficiency and reliability of the SLR are effectively improved, providing important support for scientific research and medical fields.

[0206] Those skilled in the art will appreciate that the present application may be implemented as a system, method or computer program product.

[0207] Therefore, the present disclosure may be specifically implemented in the following forms, namely: it may be completely hardware, it may be completely software (including firmware, resident software, microcode, etc.), or it may be a combination of hardware and software, generally referred to herein as a "circuit", "module" or "system". In addition, in some embodiments, the present application may also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium may contain computer-readable program code.

[0208] Any combination of one or more computer-readable media can be used. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0209] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art can change, modify, replace and modify the above embodiments within the scope of the present application. On this basis, a variety of replacements and improvements can be made to the present application, all of which fall within the scope of protection of the present application.

Claims

1. A systematic literature review method based on a large language model, characterized by: The following steps are involved: Generate a first draft of the SLR solution based on the large language model according to project requirements; Confirm the draft of the SLR plan to obtain the SLR research plan; Based on the SLR research plan, literature search and analysis were performed to obtain preliminary search results; Based on the preliminary search results, the SLR research plan was modified to obtain the literature review results.

2. The method for systematic literature review based on a large language model according to claim 1, characterized in that: The steps to generate the first draft of the SLR solution based on the large language model according to the project requirements are as follows: The SLR Protocol based on the large language model generates the relevant SLR plan draft according to the project requirements.

3. The method for systematic literature review based on a large language model according to claim 2, characterized in that: The steps of generating the relevant SLR scheme draft based on the SLR Protocol of the large language model according to the project requirements specifically include: Generate background and research rationale based on large language models; Generate specific goals based on the large language model; Generative methods based on large language models; Generate retrieval strategies based on large language models.

4. The method for systematic literature review based on a large language model according to claim 3, characterized in that: The steps to generate background and research rationale based on the large language model include: Enter the title of your study; According to the research title, analysis and context generation are performed based on a large language model; Identify and reason about research questions based on large language models; Provide complete background and rationale for your research.

5. The method for systematic literature review based on a large language model according to claim 4, characterized in that: The steps to generate specific targets based on the large language model include: Through a large model based on the Transformer architecture, relevant information in the field is extracted according to the research topic or keywords and a description of the purpose of the review is generated; Using a pre-trained language model, the research elements are analyzed based on the input research background information and the PICO / PEO framework, and a standardized expression of the main research questions is generated through semantic reasoning; Through the named entity recognition technology of the large model, the relevant features of the research population are extracted from the domain corpus, and the definition of the research population is generated by combining the contextual semantic relationship; Use deep learning-based language generation technology to parse intervention program keywords and retrieve relevant information from the database to generate standardized descriptions of intervention measures; Through semantic matching technology, the research requirements input by users are parsed based on a large language model, and comparison group options are extracted and definitions are generated; Through the indicator recommendation module based on the pre-trained large model, the input research background and objectives are analyzed, the main outcome indicators are extracted, and customized outcome indicator descriptions are generated based on user needs.

6. The method for systematic literature review based on a large language model according to claim 5, characterized in that: The steps of the large language model generation method specifically include: Using a large language model, we generate inclusion and exclusion criteria by semantically parsing the domain corpus and the research objectives input by users; The large model was used to analyze the research topics and combine the research design types to generate the included research design criteria and exclusion types; Through the large language model, combined with the medical knowledge base and semantic analysis technology, the research background is analyzed and the detailed characteristics of the research object are generated; Use large language models to extract interventions or exposures and generate structured descriptions of doses, dosing schedules, and other parameters. The domain literature and experimental design guidelines are parsed through a large language model to generate comparison group definitions; Based on a large language model and domain knowledge base, the primary and secondary outcome indicators related to the research topic are automatically extracted to generate structured content; The research needs and literature metadata are parsed through a large model to generate screening criteria for language and publication date.

7. The method for systematic literature review based on a large language model according to claim 6, characterized in that: The steps to generate a retrieval strategy based on a large language model are as follows: The literature search strategy generation technology based on the large language model is used to generate a complete search strategy.

8. The method for systematic literature review based on a large language model according to claim 7, characterized in that: The complete search strategy covers database recommendation, keyword optimization, grey literature expansion and restriction definition.

9. The method for systematic literature review based on a large language model according to claim 8, characterized in that: The formats of the SLR study protocol include PRISMA format and COCHRANE format.

10. A systematic literature review system based on a large language model, characterized in that: include: SLR draft module, used to generate SLR draft according to project requirements based on the large language model; An SLR research plan module is used to confirm the SLR plan draft and obtain the SLR research plan; A literature search and analysis module is used to search and analyze literature based on the SLR research plan to obtain preliminary search results; A scheme modification module is used to modify the SLR research scheme based on the preliminary search results to obtain literature review results.

Images