Risk assessment method, system, device and medium based on graph search enhancement generation and multi-agent collaboration mechanism

Abstract

The application relates to the technical field of artificial intelligence information processing. The application discloses a risk assessment method and system based on graph search enhanced generation and a multi-agent collaboration mechanism, equipment and a medium, which can realize automatic processing of risk assessment, reduce the workload of manual analysis and improve the accuracy of risk assessment. The method comprises the following steps: acquiring background description, reference materials and risk problems corresponding to a target disease; the reference materials are generated by constructing a disease topic knowledge graph based on graph search enhanced generation; the background description, the reference materials and the risk problems are integrated and processed to generate a pre-meeting preparation document; a plurality of simulated experts are called according to a preset discussion round to perform risk assessment processing on the pre-meeting preparation document, and a risk assessment result of the target disease is generated.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence information processing technology. More specifically, this application relates to a risk assessment method, system, device, and medium based on graph retrieval-enhanced generation and multi-agent collaborative mechanism. Background Technology

[0002] With the rapid evolution of information technology and artificial intelligence, risk assessment is playing an increasingly important role in public health, healthcare, drug regulation, and emergency monitoring. In disease-specific risk assessment applications, traditional methods primarily rely on manual collection, screening, and analysis of background materials, literature, and relevant information to arrive at a risk assessment result. This manual approach is not only time-consuming and labor-intensive but also highly dependent on expert experience. Furthermore, expert experience is inherently subjective, which can easily lead to discrepancies between the risk assessment results and reality, thus reducing the accuracy of the risk assessment. Summary of the Invention

[0003] The purpose of this application is to provide a risk assessment method, system, device, and medium based on graph retrieval-enhanced generation and multi-agent collaborative mechanisms. This method enables automated risk assessment, reduces the workload of manual analysis, and improves the accuracy of risk assessment. This application is mainly achieved through the following technical solutions: A first aspect of this application provides a risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanisms, comprising: Obtain the background description, reference materials, and risk issues corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement. The background description, reference materials, and risk issues are integrated and processed to generate a pre-meeting preparation document; Based on the preset discussion rounds, multiple simulated experts are invoked to perform risk assessment processing on the pre-meeting preparation documents, generating risk assessment results for the target disease.

[0004] According to one embodiment of this application, the step of obtaining the background description includes: Obtain the webpage link corresponding to the target disease; The web page cleaning interface is called to perform ad removal, navigation bar removal, and text extraction on the web page content corresponding to the web page link, thereby obtaining the structured text corresponding to the web page link. The structured text is processed by topic identification, summary generation, and key information extraction to obtain the background description.

[0005] According to one embodiment of this application, the step of obtaining the reference material includes: Obtain the disease-themed knowledge graph constructed based on graph retrieval enhancement; Generate vector indexes and graph structure indexes based on the aforementioned disease-themed knowledge graph; The background description is semantically encoded to obtain target encoded information; The target encoding information, the vector index, and the graph structure index are fused together to obtain enhanced contextual information. The reference data is generated based on the enhanced contextual information.

[0006] According to one embodiment of this application, the steps for obtaining the disease topic knowledge graph constructed based on graph retrieval enhancement include: Obtain multiple research papers and news information corresponding to the target disease; After text translation processing of all research literature and all news information, the target literature corresponding to each research literature and the target news corresponding to each news information are obtained, and all target literature and all target news are written into the knowledge base according to the topic of the target disease. The disease-themed knowledge graph is constructed based on all entities and relationships in the knowledge base.

[0007] According to one embodiment of this application, the step of constructing the disease-themed knowledge graph based on all entities and all relationships in the knowledge base includes: All target documents and all target news in the knowledge base are cleaned to obtain the final document corresponding to each target document and the final news corresponding to each target news item. Entity and relationship identification processing is performed on all final documents and all final news to obtain multiple target entities and multiple target relationships; The disease-themed knowledge graph is constructed based on the multiple target entities and the multiple target relationships.

[0008] According to one embodiment of this application, the step of obtaining the risk issue includes: The risk question is generated based on the key risk nodes in the disease-themed knowledge graph.

[0009] According to one embodiment of this application, the step of calling multiple simulated experts to perform risk assessment processing on the pre-meeting preparation document according to a preset discussion round, and generating a risk assessment result for the target disease, includes: Based on the preset discussion rounds, multiple simulated experts are invoked to conduct multiple rounds of discussion on the pre-meeting prepared documents to obtain roundtable discussion records; The roundtable discussion record is processed for risk assessment to generate risk assessment results for the target disease.

[0010] A second aspect of this application provides a risk assessment system based on graph retrieval-enhanced generation and multi-agent collaborative mechanisms, comprising: The information acquisition module is used to acquire background descriptions, reference materials, and risk issues corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement. The integration and processing module is used to integrate and process the background description, the reference materials, and the risk issues to generate pre-meeting preparation documents; The risk assessment module is used to call multiple simulated experts to perform risk assessment processing on the pre-meeting preparation documents according to the preset discussion rounds, and generate risk assessment results for the target disease.

[0011] A third aspect of this application provides a terminal device, including a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to perform the steps of the risk assessment method based on graph retrieval enhanced generation and multi-agent collaborative mechanism provided in the first aspect of this application.

[0012] A fourth aspect of this application provides a computer-readable storage medium for storing a computer program that causes a computer to perform the steps of the risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism provided in the first aspect of this application.

[0013] The beneficial effects of the embodiments of this application include: This application embodiment obtains background descriptions, reference materials, and risk questions corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement. The background descriptions, reference materials, and risk questions are integrated to generate a pre-meeting preparation document. Multiple simulated experts are invoked according to a preset discussion round to perform risk assessments on the pre-meeting preparation document, generating a risk assessment result for the target disease. Compared to existing technologies using manual methods, this application embodiment uses multiple simulated experts to replace manual risk assessment of the data. Therefore, this application embodiment automates risk assessment, reducing the workload and assessment time of manual analysis and significantly weakening the influence of subjective human factors on the risk assessment results, thereby improving the accuracy of risk assessment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 The flowcharts for some embodiments of the risk assessment method based on graph retrieval-enhanced generation and multi-agent collaboration mechanism of this application are shown below. Figure 2 This is a block diagram of the risk assessment system based on graph retrieval-enhanced generation and multi-agent collaboration mechanism in some embodiments of this application; Figure 3 This is a schematic block diagram of the terminal device of this application in some embodiments. Detailed Implementation

[0016] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0017] It should be noted that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0018] The terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are expressly listed, but may include other steps or units that are not expressly listed or that are inherent to such process, method, product, or apparatus.

[0020] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The term "and / or" as used in this application includes any and all combinations of one or more of the associated listed items.

[0021] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0022] refer to Figure 1 The diagram shown is a flowchart of a risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism, provided in the first aspect of an embodiment of this application. Figure 1 The risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism includes: S1. Obtain the background description, reference materials, and risk issues corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement.

[0023] Furthermore, the background description acquisition step includes: acquiring the webpage link corresponding to the target disease; calling the webpage cleaning interface to perform ad removal, navigation bar removal, and text extraction processing on the webpage content corresponding to the webpage link to obtain the structured text corresponding to the webpage link; and performing topic recognition, summary generation, and key information extraction processing on the structured text to obtain the background description.

[0024] The target disease can be selected from any one of coronary heart disease, cirrhosis, leukemia, diabetes, kidney stones, pneumonia, and Parkinson's disease. In other embodiments, the target disease can also be other diseases, which can be set by those skilled in the art according to their needs.

[0025] The webpage link is entered by the user. The webpage cleaning interface is Jina Render.

[0026] Furthermore, the steps of calling the webpage cleaning interface to perform ad removal, navigation bar removal, and text extraction processing on the webpage content corresponding to the webpage link to obtain the structured text corresponding to the webpage link include: calling the webpage cleaning interface to parse the webpage content corresponding to the webpage link into a tree structure; calculating the text density of each node in the tree structure; identifying nodes with text density lower than a preset value as nodes corresponding to ads and navigation bars, and deleting the nodes corresponding to ads and navigation bars to obtain multiple target nodes; and performing text extraction processing on the text corresponding to all target nodes to obtain the structured text corresponding to the webpage link.

[0027] The tree structure can be a DOM node tree. In other embodiments, the tree structure can also be other structures, which can be set by those skilled in the art according to actual needs.

[0028] The structured text is pure text.

[0029] The implementation of calling the webpage cleaning interface to perform ad removal, navigation bar removal, and text extraction on the webpage content corresponding to the webpage link, thereby obtaining the structured text corresponding to the webpage link, can significantly improve the quality and parsability of webpage information, provide highly reliable input for subsequent knowledge extraction and graph construction, and reduce manual preprocessing costs.

[0030] This application's embodiments introduce webpage denoising, text extraction, and semantic parsing mechanisms, and utilize Graph RAG's graph structure construction and relationship recognition capabilities to uniformly transform information from different sources into a structured knowledge graph, fundamentally solving the problem of isolated multi-source data that is difficult to directly use for reasoning and analysis.

[0031] Furthermore, the steps of performing topic recognition, summary generation, and key information extraction on the structured text to obtain the background description include: In this embodiment of the application, a locally deployed large language model is used to perform topic recognition, summary generation, and key information extraction on the structured text to obtain the background description.

[0032] The large language model can be DeepSeek V3. In other embodiments, the large language model can also be other models, which can be set by those skilled in the art according to actual needs. The large language model can be understood as a background extraction agent.

[0033] The key information includes specific descriptions and / or evaluation information of the disease. In other embodiments, the key information may also include other content, which can be set by those skilled in the art according to actual needs.

[0034] The background description provides a basic semantic framework for subsequent knowledge supplementation and graph construction.

[0035] Furthermore, the steps for obtaining the reference materials include: obtaining the disease-themed knowledge graph constructed based on graph retrieval enhancement; generating a vector index and a graph structure index based on the disease-themed knowledge graph; performing semantic encoding processing on the background description to obtain target encoding information; fusing the target encoding information, the vector index, and the graph structure index to obtain enhanced contextual information; and generating the reference materials based on the enhanced contextual information.

[0036] Furthermore, the steps for obtaining the disease-themed knowledge graph constructed based on graph retrieval enhancement include: obtaining multiple research documents and multiple news information corresponding to the target disease; performing text translation processing on all research documents and all news information to obtain the target document corresponding to each research document and the target news corresponding to each news information, and writing all target documents and all target news into the knowledge base according to the theme of the target disease; and constructing the disease-themed knowledge graph based on all entities and all relationships in the knowledge base.

[0037] Furthermore, the steps of obtaining multiple research papers and multiple news information corresponding to the target disease include: automatically crawling multiple research papers corresponding to the target disease by accessing external literature retrieval services; and continuously obtaining real-time risk events from global news platforms, including EIOS, as the multiple news information by using RSS subscription channels.

[0038] The external document retrieval service is based on the PanShi model. In other embodiments, the external document retrieval service may be other models, which can be specifically set by those skilled in the art according to actual needs. The external document retrieval service can be understood as a document retrieval interface.

[0039] The embodiments of this application can realize the dynamic, continuous and automatic collection of academic literature (i.e., the research literature), information on breaking events and global news.

[0040] Furthermore, the step of constructing the disease-themed knowledge graph based on all entities and all relationships in the knowledge base includes: performing text cleaning processing on all target documents and all target news in the knowledge base to obtain the final document corresponding to each target document and the final news corresponding to each target news; performing entity and relationship recognition processing on all final documents and all final news to obtain multiple target entities and multiple target relationships; and constructing the disease-themed knowledge graph based on the multiple target entities and the multiple target relationships.

[0041] Furthermore, the steps of performing text cleaning processing on all target documents and all target news articles in the knowledge base to obtain the final document corresponding to each target document and the final news article corresponding to each target news article include: removing headers and footers, removing reference lists, removing appendices, removing figure and table titles, and removing duplicate paragraphs from all target documents in the knowledge base to obtain the final document corresponding to each target document; and removing advertising code, removing copyright statements, removing duplicate leads, and removing reporter byline from all target news articles in the knowledge base to obtain the final news article corresponding to each target news article.

[0042] Furthermore, the step of constructing the disease topic knowledge graph based on the multiple target entities and the multiple target relationships includes: connecting the multiple target entities based on the multiple target relationships to obtain the disease topic knowledge graph.

[0043] The plurality of target entities includes diseases, symptoms, and drugs. In other embodiments, the plurality of target entities may also include other entities, which can be specifically set by those skilled in the art according to actual needs.

[0044] The multiple target relationships include "disease-symptom", "disease-drug", and "symptom-drug". In other embodiments, the multiple target relationships may also include other relationships, which can be set by those skilled in the art according to actual needs.

[0045] The disease-themed knowledge graph supports incremental updates and structured reasoning, which can significantly improve the system's knowledge completeness and interpretability.

[0046] Furthermore, the steps of generating vector indexes and graph structure indexes based on the disease-themed knowledge graph include: using a vector indexing algorithm to generate vector indexes for all entities in the disease-themed knowledge graph; and using an R-tree algorithm to perform index generation processing on all entities and all relations in the disease-themed knowledge graph to obtain the graph structure index.

[0047] The vector indexing algorithm is the FAISS algorithm. Faiss is an open-source similarity search tool developed by Facebook AI Labs (FAIR) in 2015, designed to solve the problem of fast retrieval of high-dimensional vector data and overcome the technical limitations of traditional search engines. In other embodiments, the vector indexing algorithm can also be other algorithms, which can be specifically set by those skilled in the art according to actual needs.

[0048] The fusion process can also be a fusion retrieval process.

[0049] Furthermore, the step of generating the reference material based on the enhanced contextual information includes: extracting the abstract and the main text from the enhanced contextual information to obtain the reference material.

[0050] The references mentioned are structured references.

[0051] Furthermore, the step of obtaining the risk question includes: generating the risk question based on the key risk nodes in the disease topic knowledge graph.

[0052] The critical risk node refers to the point at which a state transitions from one state to another.

[0053] Furthermore, the step of generating the risk question based on the key risk nodes in the disease topic knowledge graph includes: querying the question template for the question corresponding to the key risk node as the risk question.

[0054] This application embodiment involves multiple rounds of interactive discussions among simulated experts around key risk nodes in a Graph RAG, utilizing the evidence chain, event correlation, and knowledge structure in the graph to support the reasoning process, making the discussion results more in-depth, comprehensive, and interpretable, closely resembling the risk assessment process of a real expert meeting.

[0055] Through the above implementation methods, the embodiments of this application can simultaneously integrate heterogeneous data from multiple sources, such as background descriptions, references, and risk issues, making the evidence on which the risk assessment is based more sufficient and comprehensive.

[0056] S2. Integrate and process the background description, reference materials, and risk issues to generate a pre-meeting preparation document.

[0057] S3. Based on the preset discussion rounds, call multiple simulated experts to perform risk assessment processing on the pre-meeting preparation documents and generate risk assessment results for the target disease.

[0058] The specific values ​​for the preset discussion rounds can be set by those skilled in the art according to actual needs. The preset discussion rounds can be understood as rounds of roundtable discussion.

[0059] The preset discussion rounds are input by the user.

[0060] Furthermore, step S3 includes: calling multiple simulated experts to conduct multiple rounds of discussion on the pre-meeting preparation document according to a preset discussion round, and obtaining roundtable discussion records; performing risk assessment processing on the roundtable discussion records to generate risk assessment results for the target disease.

[0061] The preset discussion rounds are set by those skilled in the art according to actual needs.

[0062] Through the above implementation methods, this application embodiment achieves a cyclical discussion mode of viewpoint presentation, evidence citation, logical verification, and conclusion formation by having multiple simulated experts conduct multi-round reasoning discussions based on the evidence chain provided by Graph RAG. This mechanism can simulate the knowledge collision process in real risk assessment meetings, thereby improving the depth of analysis and the reliability of assessment conclusions.

[0063] This application embodiment calls multiple simulated experts to replace manual methods for risk assessment of data. Thus, this application embodiment realizes the automated processing of risk assessment, which can not only reduce the workload and assessment time of manual analysis, but also greatly weaken the influence of human subjective factors on the risk assessment results, thereby improving the accuracy of risk assessment.

[0064] In addition, the embodiments of this application can also solve the problems that have long existed in the prior art in risk assessment scenarios, such as process fragmentation, lagging knowledge updates and lack of professional analysis capabilities.

[0065] In some implementations, the pre-meeting preparation document may also be generated by integrating the background description, the reference materials, the risk issues, the background text, and the name of the target disease.

[0066] Both the background text and the name of the target disease are input by the user.

[0067] In some implementations, after performing entity and relation identification processing on all final documents and all final news to obtain multiple target entities and multiple target relations, the step of constructing the disease topic knowledge graph based on all entities and all relations in the knowledge base further includes: vectorizing all the multiple target entities and multiple target relations to obtain the final entity corresponding to each target entity and the final relation corresponding to each target relation; and constructing the disease topic knowledge graph based on all final entities and all final relations.

[0068] In some embodiments, the risk assessment method based on graph retrieval-enhanced generation and multi-agent collaboration mechanism further includes invoking a role-generating agent to generate the multiple simulated experts, wherein the multiple simulated experts have different professional backgrounds, including research directions, viewpoint foundations, and academic characteristics. In other embodiments, the professional backgrounds may also include other backgrounds, which can be specifically set by those skilled in the art according to actual needs.

[0069] The role-generating intelligent agent can be understood as a generative agent, which is an intelligent agent that simulates human behavior based on a generative model.

[0070] Through the above implementation methods, this application's embodiments do not require manual pre-definition of expert roles. Instead, they automatically generate diverse expert profiles (i.e., simulated experts) based on specific disease themes and knowledge graph content. Each expert profile includes corresponding professional background, academic stance, research direction, and reasoning style. This mechanism can simulate the composition of a real expert group, achieving comprehensive reasoning capabilities from multiple perspectives and across multiple fields.

[0071] The operating environment of this application embodiment can adopt a portable Conda environment. The Conda environment allows computational tasks, including graph construction, Graph RAG (graph retrieval enhanced generation) inference, multi-agent collaboration, and text generation, to be executed independently locally, effectively meeting the needs of data privacy protection, secure isolation deployment, and cross-device migration. Through the above design, this invention constructs a complete, automated, portable, and scalable intelligent risk assessment process, providing reliable technical support for scenarios such as public health monitoring, medical assessment, drug regulation, and emergency event analysis. This application embodiment uses a locally deployed large model and a portable Conda environment, enabling the entire process of data acquisition, knowledge extraction, graph updating, and Graph RAG inference to run locally. This solution protects the security of sensitive data, improves system reliability, and provides stronger controllability and customizability for specific industry applications.

[0072] Graph RAG technology is used to construct entity nodes, attribute nodes, and relation edges to generate a disease-themed knowledge graph that can be used for graph-structured reasoning. The embodiments of this application can simultaneously construct a composite retrieval index that combines embedded vectors and graph structures, providing support for multimodal reasoning.

[0073] This application embodiment constructs a disease-themed knowledge graph using Graph RAG. By automatically extracting disease-related entities, attributes, and semantic relationships, it forms a disease knowledge network with structured features. Combined with the disease research directions generated by the intelligent agent, this application embodiment can adaptively adjust risk focus, background organization strategies, and analysis logic to meet the specific risk assessment needs of different diseases, and has a high degree of specialization and scalability.

[0074] The embodiments of this application also support rapid migration and deployment across devices to reduce maintenance and expansion costs, and are suitable for fields with strict data security requirements, such as healthcare.

[0075] This application's embodiments introduce Graph RAG as a unified knowledge enhancement engine in the multi-agent collaborative work process. It not only uses vector retrieval to provide semantically similar content but also utilizes entity relationships, association paths, and evidence chains in the graph to achieve structured retrieval. This mechanism enables different agents to share a unified, structured knowledge base when performing reference generation, risk problem analysis, and discussion reasoning, thereby significantly improving information consistency, reasoning stability, and result accuracy.

[0076] refer to Figure 2 The diagram shown is a principle block diagram of a risk assessment system based on graph retrieval-enhanced generation and multi-agent collaborative mechanism, provided in the second aspect of an embodiment of this application. Figure 2The risk assessment system 100 based on graph retrieval-enhanced generation and multi-agent collaborative mechanism includes: The information acquisition module 101 is used to acquire the background description, reference materials and risk issues corresponding to the target disease. The reference materials are generated by a disease topic knowledge graph constructed based on graph retrieval enhancement. The integration processing module 102 is used to integrate and process the background description, the reference materials and the risk issues to generate a pre-meeting preparation document; The risk assessment module 103 is used to call multiple simulated experts to perform risk assessment processing on the pre-meeting preparation documents according to the preset discussion rounds, and generate the risk assessment results for the target disease.

[0077] The integrated processing module 102 can be understood as a pre-meeting document generation module, which can automatically construct structured pre-meeting preparation materials, including background overview, evidence chain, and core issue list.

[0078] In some embodiments, the information acquisition module 101 includes a parsing module and a background extraction agent module. The parsing module is used to acquire the webpage link corresponding to the target disease; call the webpage cleaning interface to perform ad removal, navigation bar removal, and text extraction processing on the webpage content corresponding to the webpage link to obtain the structured text corresponding to the webpage link; the background extraction agent module is used to perform topic recognition, summary generation, and key information extraction processing on the structured text to obtain the background description.

[0079] In some embodiments, the information acquisition module 101 further includes a Graph RAG retrieval enhancement module and a reference material generation agent module. The Graph RAG retrieval enhancement module is used to acquire the disease-themed knowledge graph constructed based on the graph retrieval enhancement; generate vector indexes and graph structure indexes based on the disease-themed knowledge graph; perform semantic encoding processing on the background description to obtain target encoding information; and fuse the target encoding information, the vector indexes, and the graph structure indexes to obtain enhanced contextual information. The reference material generation agent module is used to generate the reference materials based on the enhanced contextual information.

[0080] The Graph RAG retrieval enhancement module not only supports traditional vector matching, but also performs structured retrieval based on graph node relationships, associated paths and semantic chains, thereby generating highly relevant and interpretable contextual material as input for downstream agents.

[0081] The reference data generation intelligent agent module relies on the high-quality search materials provided by Graph RAG to automatically generate systematic reference data related to the current disease or risk topic.

[0082] In some implementations, the Graph RAG retrieval enhancement module includes a literature and news acquisition module and a knowledge base construction module. The literature and news acquisition module is used to acquire multiple research documents and multiple news information corresponding to the target disease. The knowledge base construction module is used to perform text translation processing on all research documents and all news information to obtain the target document corresponding to each research document and the target news corresponding to each news information, and write all target documents and all target news into the knowledge base according to the topic of the target disease. Based on all entities and all relationships in the knowledge base, the disease topic knowledge graph is constructed.

[0083] In some implementations, the literature and news acquisition module includes a literature acquisition submodule and a news gathering submodule, wherein the literature acquisition submodule is used to automatically retrieve multiple research documents corresponding to the target disease by accessing external literature retrieval services; and the news gathering submodule is used to continuously acquire real-time risk events from global news platforms, including EIOS, as the multiple news information by using RSS subscription channels.

[0084] In some embodiments, the information acquisition module 101 further includes a risk question generation agent module, which is used to generate the risk question based on key risk nodes in the disease topic knowledge graph.

[0085] The risk issue generation intelligent agent module automatically generates core issues such as epidemic trends, transmission mechanisms, clinical risks, public health risks, and regulatory risks around key risk nodes in the graph, laying a structured topic framework for subsequent discussions.

[0086] In some implementations, the risk assessment module 103 includes a roundtable discussion simulation module and an assessment module. The roundtable discussion simulation module is used to call multiple simulated experts to conduct multiple rounds of discussion on the pre-meeting preparation document according to a preset number of discussion rounds, and obtain roundtable discussion records. The assessment module is used to perform risk assessment processing on the roundtable discussion records and generate risk assessment results for the target disease.

[0087] The roundtable discussion simulation module is used to activate multiple agents, each representing a different expert role, to conduct multiple rounds of discussions on risk issues from their respective professional perspectives. This embodiment of the application can track each round of speeches, reasoning paths, evidence connections, and the resulting interim conclusions, achieving an interpretable collaborative reasoning process and ultimately generating a comprehensive discussion result with an expert conference style.

[0088] In some implementations, the risk assessment system based on graph retrieval-enhanced generation and multi-agent collaboration mechanism further includes an expert role generation module, which calls a role generation agent to generate the multiple simulated experts.

[0089] The expert role generation module automatically constructs virtual expert roles participating in roundtable discussions using a large model, including information such as their research field, professional characteristics, work background, and academic stance. This module ensures that each role is differentiated and independent to simulate the multi-perspective structure of a real expert meeting.

[0090] A third aspect of this application provides a terminal device, the schematic diagram of which is as follows: Figure 3 As shown. The terminal device includes a processor, memory, network interface, display screen, and temperature sensor connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the terminal device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanisms. The display screen can be a liquid crystal display (LCD) or an e-ink display. The temperature sensor is pre-installed inside the terminal device to detect the operating temperature of the internal components.

[0091] Those skilled in the art will understand that Figure 3 The schematic diagram shown is only a partial structural diagram related to the present invention and does not constitute a limitation on the terminal device to which the present invention is applied. The specific terminal device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0092] In some embodiments, this application provides a terminal device, which includes a processor and a memory for storing computer programs. The processor is used to call and run the computer programs stored in the memory to perform the steps of the risk assessment method based on graph retrieval enhancement generation and multi-agent collaborative mechanism provided in the first aspect of this application.

[0093] A fourth aspect of this application provides a computer-readable storage medium for storing a computer program that causes a computer to perform the steps of the risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism provided in the first aspect of this application.

[0094] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0095] The technical features of the above embodiments can be combined without changing the basic principles of this application. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0096] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the patent protection scope of this application should be determined by the appended claims.

Claims

1. A risk assessment method based on graph search enhancement generation and multi-agent collaborative mechanism, characterized in that, include: Obtain the background description, reference materials, and risk issues corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement. The background description, reference materials, and risk issues are integrated and processed to generate a pre-meeting preparation document; Based on the preset discussion rounds, multiple simulated experts are invoked to perform risk assessment processing on the pre-meeting preparation documents, generating risk assessment results for the target disease.

2. The method of claim 1, wherein the risk assessment is generated based on a graph search enhancement and a multi-agent collaboration mechanism. The steps for obtaining the background description include: Obtain the webpage link corresponding to the target disease; The web page cleaning interface is called to perform ad removal, navigation bar removal, and text extraction on the web page content corresponding to the web page link, thereby obtaining the structured text corresponding to the web page link. The structured text is processed by topic identification, summary generation, and key information extraction to obtain the background description.

3. The method of claim 1, wherein the method is based on a graph search enhanced generation of risk assessment with multi-agent coordination mechanism. The steps for obtaining the reference materials include: Obtain the disease-themed knowledge graph constructed based on graph retrieval enhancement; Generate vector indexes and graph structure indexes based on the aforementioned disease-themed knowledge graph; The background description is semantically encoded to obtain target encoded information; The target encoding information, the vector index, and the graph structure index are fused together to obtain enhanced contextual information. The reference data is generated based on the enhanced contextual information.

4. The risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism according to claim 3, characterized in that, The steps for obtaining the disease-themed knowledge graph constructed based on graph retrieval enhancement include: Obtain multiple research papers and news information corresponding to the target disease; After text translation processing of all research literature and all news information, the target literature corresponding to each research literature and the target news corresponding to each news information are obtained, and all target literature and all target news are written into the knowledge base according to the topic of the target disease. The disease-themed knowledge graph is constructed based on all entities and relationships in the knowledge base.

5. The risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism according to claim 4, characterized in that, The steps for constructing the disease-themed knowledge graph based on all entities and relationships in the knowledge base include: All target documents and all target news in the knowledge base are cleaned to obtain the final document corresponding to each target document and the final news corresponding to each target news item. Entity and relationship identification processing is performed on all final documents and all final news to obtain multiple target entities and multiple target relationships; The disease-themed knowledge graph is constructed based on the multiple target entities and the multiple target relationships.

6. The risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism according to claim 1, characterized in that, The steps for identifying the aforementioned risk issues include: The risk question is generated based on the key risk nodes in the disease-themed knowledge graph.

7. The risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism according to claim 1, characterized in that, The steps for invoking multiple simulated experts to perform risk assessments on the pre-meeting prepared documents according to preset discussion rounds, and generating risk assessment results for the target disease, include: Based on the preset discussion rounds, multiple simulated experts are invoked to conduct multiple rounds of discussion on the pre-meeting prepared documents to obtain roundtable discussion records; The roundtable discussion record is processed for risk assessment to generate risk assessment results for the target disease.

8. A risk assessment system based on graph retrieval-enhanced generation and multi-agent collaborative mechanism, characterized in that, include: The information acquisition module is used to acquire background descriptions, reference materials, and risk issues corresponding to the target disease. The reference materials are generated from a disease-themed knowledge graph constructed based on graph retrieval enhancement. The integration and processing module is used to integrate and process the background description, the reference materials, and the risk issues to generate pre-meeting preparation documents; The risk assessment module is used to call multiple simulated experts to perform risk assessment processing on the pre-meeting preparation documents according to the preset discussion rounds, and generate risk assessment results for the target disease.

9. A terminal device, characterized in that, include: A processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to perform the steps of the risk assessment method based on graph retrieval enhancement generation and multi-agent collaborative mechanism as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program that causes a computer to perform the steps of the risk assessment method based on graph retrieval-enhanced generation and multi-agent collaborative mechanism as described in any one of claims 1 to 7.

Images