Multi-agent reasoning system, data reasoning method, apparatus and device, and product

Abstract

Provided in the present application are a multi-agent reasoning system, a data reasoning method, apparatus and device, and a product. The method is applied to a multi-agent reasoning system comprising a plurality of agents, and comprises: a first agent performing key information extraction on reasoning reference data on the basis of a reasoning problem, so as to obtain reasoning key information (S101); a second agent generating a reasoning path on the basis of the reasoning problem and the reasoning key information (S102), wherein the reasoning path comprises prompt information for performing reasoning on the basis of the reasoning key information, so as to solve the reasoning problem; and a third agent performing integrated reasoning on the basis of the reasoning problem and the reasoning path, so as to generate a reasoning result corresponding to the reasoning problem (S103), wherein the reasoning result comprises a reasoning process and / or a reasoning conclusion. The method can improve the flexibility and reasoning efficiency of data reasoning.

Description

Multi-agent reasoning system, data reasoning method, device, equipment and product TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a multi-agent reasoning system, a data reasoning method, a device, equipment and product. BACKGROUND

[0002] Data reasoning refers to reasoning processing under certain direction or certain demand for specific data, so as to obtain a reasoning conclusion. Data reasoning is a common means of extracting or understanding important information from data, and making decisions based on data. For example, in the medical auxiliary diagnosis scene, it is often necessary to comprehensively reason and analyze the medical records, examination reports, physical examination reports and the like of patients, so as to determine the diseases and corresponding diagnosis and treatment schemes of the patients.

[0003] Traditional data reasoning is usually performed by professional personnel by reading, summarizing and understanding reasoning reference data, and then reasoning to obtain a reasoning result based on experience. This data reasoning method is low in efficiency, is greatly affected by the professional level of personnel, and is poor in reasoning flexibility. SUMMARY

[0004] Based on the above technical status, the present application proposes a multi-agent reasoning system, a data reasoning method, a device, equipment and product, which can improve the flexibility and reasoning efficiency of data reasoning.

[0005] The first aspect of the present application proposes a data reasoning method applied to a multi-agent reasoning system including a plurality of agents, the method comprising: a first agent extracting key information from reasoning reference data according to a reasoning problem, to obtain reasoning key information; a second agent generating a reasoning path according to the reasoning problem and the reasoning key information; the reasoning path comprising prompt information for reasoning based on the reasoning key information to answer the reasoning problem; and a third agent integrating reasoning according to the reasoning problem and the reasoning path, to generate a reasoning result corresponding to the reasoning problem, the reasoning result comprising a reasoning process and / or a reasoning conclusion.

[0006] In some implementations, the method further comprises: the third agent evaluating the reasoning result to obtain an evaluation conclusion; in a case where the evaluation conclusion represents that the reasoning result is incorrect, the third agent calling the second agent to regenerate the reasoning path, and the third agent integrating reasoning according to the reasoning problem and the reasoning path regenerated by the second agent, to regenerate the reasoning result corresponding to the reasoning problem.

[0007] In some implementations, in the process of the second agent generating the reasoning path according to the reasoning problem and the reasoning key information, a search question is also generated, the search question being used to search out knowledge required for executing the reasoning path from a knowledge base.

[0008] In some implementations, the method further includes: the fourth agent retrieving, according to the retrieval question, a retrieval result corresponding to the retrieval question from the knowledge base; the retrieval result including relevant knowledge corresponding to the retrieval question, the knowledge base being constructed by the fifth agent.

[0009] In some implementations, the third agent performs integrated reasoning according to the reasoning question and the reasoning path to generate a reasoning result corresponding to the reasoning question, including: the third agent performing integrated reasoning according to the reasoning question, the retrieval result, and the reasoning path to generate the reasoning result corresponding to the reasoning question.

[0010] In some implementations, the fifth agent constructs the knowledge base by performing the following processing: performing knowledge segmentation on the knowledge source file to obtain a knowledge label; generating a knowledge triple based on the knowledge label, the knowledge triple including a knowledge entity, attribute information of the knowledge entity, and a source of the knowledge entity and the attribute information of the knowledge entity; and performing integrated processing on each knowledge triple with the knowledge entity as the center to obtain a knowledge graph.

[0011] In some implementations, the attribute information of the knowledge entity includes an encoding of chapter content in which the knowledge entity and the attribute information of the knowledge entity are located in a source file of the knowledge entity and the attribute information of the knowledge entity.

[0012] In some implementations, the retrieval result further includes evaluation information of the relevant knowledge.

[0013] In some implementations, the fourth agent further determines whether the retrieval question needs to be split; in a case where it is determined that the retrieval question needs to be split, the fourth agent calls the second agent to split the retrieval question into multiple retrieval sub-questions; and the fourth agent retrieves, according to the multiple retrieval sub-questions, retrieval results corresponding to each retrieval sub-question from the knowledge base respectively.

[0014] In some implementations, the reasoning question includes a diagnosis and treatment reasoning question, and the reasoning reference data includes a medical document of a patient.

[0015] In some implementations, the medical document of the patient includes multiple medical documents of the patient at different time periods; the first agent extracts, according to the reasoning question, key information from the reasoning reference data to obtain reasoning key information, including: the first agent extracts, according to the reasoning question, patient portrait information from the multiple medical documents to generate a basic information portrait and a time sequence medical portrait corresponding to the patient.

[0016] The second aspect of the present application proposes a multi-agent reasoning system including multiple agents, the multiple agents in the multi-agent system being configured to implement the data reasoning method as in the first aspect or any implementation manner of the first aspect.

[0017] The third aspect of the present application proposes another data reasoning method, comprising: extracting key information from the reasoning reference data according to the reasoning question to obtain reasoning key information; generating a reasoning path according to the reasoning question and the reasoning key information; the reasoning path comprises prompt information for reasoning based on the reasoning key information to answer the reasoning question; integrating reasoning according to the reasoning question and the reasoning path to generate a reasoning result corresponding to the reasoning question, the reasoning result comprising a reasoning process and / or a reasoning conclusion.

[0018] The fourth aspect of the present application proposes a data reasoning device, which is applied to a multi-agent reasoning system comprising a plurality of agents, and the device comprises: an information integration unit configured to extract key information from the reasoning reference data according to the reasoning question by a first agent; an analysis reasoning unit configured to generate a reasoning path according to the reasoning question and the reasoning key information by a second agent; the reasoning path comprises prompt information for reasoning based on the reasoning key information to answer the reasoning question; an integrated reasoning unit configured to integrate reasoning according to the reasoning question and the reasoning path by a third agent to generate a reasoning result corresponding to the reasoning question, the reasoning result comprising a reasoning process and / or a reasoning conclusion.

[0019] The fifth aspect of the present application proposes a data reasoning device, comprising: an input device, an output device and a multi-agent reasoning system; the input device is connected with the processor and is configured to input the reasoning question and the reasoning reference data to the processor; the output device is connected with the processor and is configured to output the reasoning result generated by the processor; the multi-agent reasoning system is configured to execute the data reasoning method as in the first aspect or any implementation manner of the first aspect.

[0020] The sixth aspect of the present application proposes a computer program product comprising computer program instructions, which, when executed by a plurality of agents in a multi-agent reasoning system, cause the multi-agent reasoning system to execute the data reasoning method as in the first aspect or any implementation manner of the first aspect.

[0021] The data reasoning method proposed in the present application completes the data reasoning process through the cooperation of different agents in the multi-agent reasoning system. In the multi-agent reasoning system, each agent automatically performs reasoning operations in different stages of data reasoning, thereby realizing fully automated data reasoning. Moreover, the multi-agent reasoning system automatically performs corresponding data reasoning operations based on the input reasoning question and reasoning reference data, without the need to specify reasoning rules, and its data reasoning is more flexible, which can meet various data reasoning needs.

[0022] In addition, based on the wisdom of each agent, each agent can focus on processing of different reasoning stages, thereby ensuring the reasoning accuracy of each reasoning link and the accuracy of the entire reasoning process and reasoning result. The reasoning process can also be included in the reasoning result, thereby making the data reasoning conclusion more explainable, enabling users to more intuitively and efficiently understand the origin of the reasoning conclusion, and improving the information acquisition efficiency of users. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0024] FIG. 1 is a flowchart of a data reasoning method according to an embodiment of the present application.

[0025] FIGS. 2-6 are flowcharts of other data reasoning methods according to embodiments of the present application.

[0026] FIG. 7 is a structural diagram of a data reasoning apparatus according to an embodiment of the present application.

[0027] FIG. 8 is a structural diagram of a data reasoning device according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] The technical solutions of the embodiments of the present application are suitable for data reasoning application scenarios, specifically, are suitable for application scenarios of reasoning given reasoning reference data according to a reasoning problem to answer the reasoning problem. The technical solutions of the embodiments of the present application can realize efficient and accurate data reasoning, and make the reasoning conclusion more explainable.

[0029] Traditional data reasoning is usually performed by professional personnel through reading, summarizing, and understanding the reasoning reference data, and then combining experience to reason to obtain a reasoning result. This data reasoning method is low in efficiency, is greatly affected by the professional level of personnel, and is low in explainability of the reasoning conclusion.

[0030] For example, in a medical diagnosis scenario, a patient will generate a large amount of medical documents such as test reports, examination reports, medical records, etc. during the medical treatment process. Through these medical documents combined with professional medical knowledge, the disease of the patient can be determined, and the diagnosis and treatment plan can be determined. At present, the disease diagnosis and treatment reasoning based on the above medical documents is mainly executed by professional doctors, that is, the doctor determines the disease and the diagnosis and treatment plan of the patient by reading various medical documents of the patient. However, manual data reasoning will bring great work pressure to the doctor, and with the enrichment and deepening of various examination items, more and more medical documents need to be checked by the doctor, thereby further reducing the diagnosis and treatment efficiency, and the above diagnosis and treatment method is greatly affected by the doctor's level, and for the patient, he can only know the final diagnosis conclusion, and has no idea about the diagnosis process and basis.

[0031] In recent years, artificial intelligence technology has developed rapidly, bringing new opportunities for data reasoning. For example, in the medical field, more and more intelligent diagnosis schemes have emerged.

[0032] However, the current intelligent reasoning scheme can only reason under specific rules for a single data file. For example, in the medical field, the intelligent diagnosis algorithm usually pre-designs diagnosis or reasoning rules for a certain type of medical document. When the content of the medical document meets the corresponding diagnosis or reasoning rule, the diagnosis conclusion corresponding to the diagnosis or reasoning rule is directly output. For the user's personalized reasoning demand or for more diverse and larger data, the current intelligent reasoning scheme is still difficult to realize integrated reasoning, and cannot effectively meet the user's demand, and can only be used as an auxiliary tool for manual reasoning.

[0033] In view of the above technical problems, a new reasoning system is proposed, and a new data reasoning method based on the reasoning system is proposed. Based on the above reasoning system and reasoning method, more efficient and intelligent data reasoning can be realized, and the reasoning conclusion has stronger explainability.

[0034] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0035] The embodiments of the present application first propose a multi-agent reasoning system, which includes a plurality of agents.

[0036] The agent refers to an agent that can perceive the environment and take actions to achieve a specific goal. It can be software, hardware or a system, with autonomy, adaptability and interaction ability. The agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions according to the knowledge and algorithms learned by itself, and then performs actions to affect the environment or achieve the predetermined goal.

[0037] The embodiments of the present application set multiple agents in the reasoning system, which are respectively used to perform different tasks, so that the whole reasoning system can realize more flexible, efficient and accurate data reasoning.

[0038] In some embodiments, the multiple agents described above can be independently set agents, for example, each agent can run independently in an independent hardware environment, and the agents are connected with each other for communication, so that the calling and cooperative interaction between different agents can be realized. For example, each agent is set in a server, a computer, a cloud server or other different execution environments, and each agent is connected through wireless network or wired network for communication, thereby forming a multi-agent reasoning system.

[0039] In other embodiments, the multiple agents described above can also be integrated in the same hardware environment, for example, the multiple agents are integrated in the same cloud server, at this time, each agent can be a different functional module in the cloud server, and each agent can share the hardware resources of the cloud server and realize cooperative interaction with each other.

[0040] In the embodiments of the present application, each agent in the multi-agent reasoning system is configured to perform different sub-tasks in data reasoning, so that multiple agents cooperate to realize the whole process of data reasoning.

[0041] For any agent in the multi-agent reasoning system, it can be composed of a single deep learning model or multiple deep learning models, and the deep learning model includes but is not limited to a generative large language model. The deep learning models in the same agent are aggregated by MOE (Mixture of Experts, integrated learning, etc.), so as to realize the function of the agent.

[0042] The structure and training method of each agent in the multi-agent reasoning system can be the same or different, and the embodiments of the present application are not limited.

[0043] Further, embodiments of the present application do not limit the type, number and training method of the deep learning model contained in each agent, in subsequent embodiments, only through exemplary embodiments, the functions of each agent and the interactive cooperation mode of each agent in the data reasoning process are introduced, so as to realize the exemplary introduction of the functions of the multi-agent reasoning system and the processing process of the data reasoning method proposed in the present application. The implementation of the function of each agent and the interactive cooperation between each agent is mainly realized by training each agent. For specific agent training methods, please refer to the conventional neural network model training method, and the embodiments of the present application will not be described in detail.

[0044] Based on the above multi-agent reasoning system, another embodiment of the present application further proposes a data reasoning method implemented in the multi-agent reasoning system, which realizes more efficient, more flexible and more accurate data reasoning through the interaction and cooperation between each agent in the multi-agent reasoning system.

[0045] In the following, the structure and function of the multi-agent reasoning system proposed in the present application and the processing process of the data reasoning method proposed in the present application will be introduced by a series of embodiments in the application scene of medical diagnosis reasoning. Through these introductions, those skilled in the art can further clarify the structure and function of the multi-agent reasoning system proposed in the present application, and at the same time, clarify the specific processing process of the data reasoning method applied to the above multi-agent reasoning system proposed in the present application.

[0046] In some embodiments, the above multi-agent reasoning system includes a first agent, a second agent and a third agent. The first agent is used to perform data integration and information extraction on the reasoning reference data, the second agent is used to generate a reasoning path for solving the reasoning problem based on the information extracted by the first agent, and the third agent is used to perform final integrated reasoning based on the reasoning path generated by the second agent and generate a reasoning result.

[0047] Based on the multi-agent reasoning system, referring to FIG. 1, the data reasoning method proposed in the embodiments of the present application includes the following steps:

[0048] S101, the first agent extracts key information from the reasoning reference data according to the reasoning problem, and obtains reasoning key information.

[0049] The reasoning reference data refers to data objects that need to be analyzed and mined for deep knowledge and information, such as in a medical diagnosis scenario, various medical documents of a patient need to be combined to determine the disease and treatment plan of the patient, at this time, the various medical documents of the patient are the reasoning reference data, and the disease and treatment plan of the patient can be determined by analyzing the reasoning reference data.

[0050] The reasoning reference data can be in any data format or any data content, such as electronic format data stored in a memory, such as text data, image data, audio and video data, and specific examples include books, articles, reports, etc.

[0051] The reasoning reference data can be generated and stored directly in a computer system, or can be obtained by recognizing physical information carriers such as books, articles, and reports.

[0052] The reasoning problem refers to the reasoning demand or purpose of reasoning the reasoning reference data. The reasoning problem represents the reasoning direction, demand, and reasoning limitation condition of the user who wants to reason the reasoning reference data.

[0053] When the user calls the multi-agent reasoning system of the present application to perform data reasoning, the reasoning problem and the reasoning reference data are first input into the multi-agent reasoning system, and the first agent in the system receives the reasoning problem and the reasoning reference data, and extracts key information from the reasoning reference data according to the reasoning problem to obtain reasoning key information.

[0054] The first agent extracts key information from the reasoning reference data, mainly de-duplicates, integrates, extracts, and induces the reasoning reference data, and extracts information useful for solving the reasoning problem from the reasoning reference data to obtain reasoning key information.

[0055] In some embodiments, in a medical diagnosis scenario, a user can input a diagnosis and treatment reasoning problem and medical documents of a patient into the multi-agent reasoning system to determine the disease of the patient.

[0056] In the above scenario, the diagnosis and treatment reasoning problem is the reasoning problem input into the reasoning system, such as "what disease do I have". The medical documents of the patient are the reasoning reference data input into the reasoning system, and the first agent extracts key information from the medical documents of the patient according to the diagnosis and treatment reasoning problem, mainly extracting information useful for solving the diagnosis and treatment reasoning problem.

[0057] The medical documents of the patient can be medical records, physical examination reports, test results, patient self-reports, etc. Moreover, the medical documents of the patient can be one medical document or multiple medical documents, for example, multiple medical documents formed by the patient at different periods of the medical treatment process.

[0058] In the case where the medical documents of the patient include multiple medical documents of the patient at different periods, when the first intelligent agent extracts key information from the multiple medical documents of the patient according to the diagnosis and treatment reasoning problem, the first intelligent agent first classifies the medical documents according to the evolution process of the patient's condition, and then integrates, extracts, and induces the classified medical documents to extract patient portrait information as reasoning key information for the subsequent reasoning process.

[0059] Among them, for the basic information of the patient's age, gender, family history, etc. that does not change in the patient's medical treatment process, the first intelligent agent extracts it to form the patient's basic information portrait.

[0060] For the patient's clinical manifestations, examination results, surgery, medication, disease, medical history, etc. at different stages, the first intelligent agent integrates and induces them into a series of time-stamped time sequence medical portraits.

[0061] For example, the first intelligent agent can obtain the following patient portrait information by extracting key information from the multiple medical documents of the patient:

[0062] Complaint: acid reflux, heartburn for half a month.

[0063] Present illness: The patient has had acid reflux and heartburn for half a month, which is more obvious after meals, accompanied by retrosternal pain, no vomiting, no nausea, poor appetite, and came to the hospital today.

[0064] Past history: The patient has been in good health.

[0065] Physical examination: clear mind, normal lung breath sounds, tenderness (+) under xiphoid process, no rebound pain, no muscle guard, and mobile dullness.

[0066] (-), normal bowel sounds.

[0067] Test:

[0068] 2020-03-08 Gastroscopy: Esophageal mucosa with mild damage, follow-up.

[0069] Age: 57 years old

[0070] Gender: male

[0071] The above basic information portrait and time sequence medical portrait jointly constitute patient portrait information, wherein the basic information portrait can represent basic information of the patient, and based on these information, the physical quality, basic physical condition of the patient, and diagnosis of some diseases with age change can be determined; the time sequence medical portrait embodies the disease characteristics, disease evolution process, development trend, and based on these information, the disease of the patient and the disease severity can be determined by systematically grasping the disease of the patient through the whole disease course.

[0072] It can be understood that the first intelligent agent extracts key information from the reasoning reference data according to the reasoning problem, can eliminate information useless for solving the reasoning problem from the reasoning reference data, can extract information useful for solving the reasoning problem from the reasoning reference data through integration and induction of the reasoning reference data, and can arrange the useful information to make the obtained reasoning key information more convenient for reasoning, thereby being beneficial to solving the reasoning problem and providing a data basis for the subsequent data reasoning process.

[0073] S102, the second intelligent agent generates a reasoning path according to the reasoning problem and the reasoning key information.

[0074] Specifically, after the first intelligent agent extracts the reasoning key information from the reasoning reference data, the reasoning key information is sent to the second intelligent agent together with the reasoning problem, and at this time, the second intelligent agent is triggered to execute the subsequent reasoning process.

[0075] After receiving the reasoning problem and the reasoning key information sent by the first intelligent agent, the second intelligent agent further analyzes and reasons the reasoning problem and the reasoning key information, thereby generating a reasoning path. The reasoning path generated by the second intelligent agent can include one reasoning path or multiple reasoning paths, and these reasoning paths can be used to obtain a reasoning result corresponding to the reasoning problem.

[0076] The reasoning path includes prompt information for reasoning based on the reasoning key information to answer the reasoning problem, that is, prompt information for determining the reasoning conclusion corresponding to the reasoning problem or the reasoning conclusion corresponding to a sub-problem in the reasoning problem based on which information in the reasoning key information.

[0077] For example, given the patient profile information provided by the first intelligent agent: "Chief complaint: acid reflux and heartburn for half a month. Present illness: The patient has experienced acid reflux and heartburn for the past half month, which is more pronounced after meals, accompanied by retrosternal pain, no vomiting, no nausea, and poor appetite. He came for treatment today. Past medical history: The patient was previously healthy. Physical examination: The patient is alert, with clear breath sounds in both lungs, tenderness below the xiphoid process (+), no rebound tenderness, no muscle guarding, shifting dullness (-), and normal bowel sounds. Laboratory examination: Gastroscopy on 2020-03-08: mild damage to the esophageal mucosa, follow-up. Age: 57 years old, gender: male," and the diagnostic reasoning question sent by the first intelligent agent: "What disease do I have?", the second intelligent agent reasoned based on this diagnostic reasoning question and generated a reasoning path from the above patient profile information.

[0078] For example, the reasoning path generated by the second agent is: "First, based on the patient's chief complaint and present medical history, the patient experiences acid reflux and heartburn, especially after meals. There is also retrosternal pain. This suggests we should consider an upper gastrointestinal disease." Based on this reasoning path, a preliminary diagnostic conclusion for the patient's disease can be drawn.

[0079] S103, the third agent integrates reasoning based on the reasoning problem and reasoning path, and generates the reasoning result corresponding to the reasoning problem.

[0080] The reasoning result includes the reasoning process and / or the reasoning conclusion.

[0081] Specifically, after generating the reasoning path, the second agent sends the generated reasoning path and reasoning question to the third agent, triggering the third agent to execute the subsequent reasoning process.

[0082] The third agent integrates reasoning based on the received reasoning question and reasoning path. For example, it integrates the reasoning basis and prompts in the reasoning path to form a reasoning process and obtain the corresponding reasoning conclusion. Alternatively, when there are multiple reasoning paths, the third agent integrates the reasoning process and reasoning conclusion of each reasoning path separately to finally obtain the reasoning result corresponding to the above reasoning question. The form of the reasoning result conforms to the answer form of the reasoning question, and the reasoning result includes the reasoning process and / or reasoning conclusion.

[0083] Including the reasoning process in the reasoning result can make the reasoning result more interpretable, allowing users to understand the origin of the reasoning conclusion more intuitively and enhancing the readability of the reasoning result.

[0084] As can be seen from the above description, the data reasoning method proposed in the embodiments of the present application completes the data reasoning process through the cooperation of different agents in the multi-agent reasoning system. In the multi-agent reasoning system, each agent automatically performs the reasoning operation in different stages of the data reasoning, so that fully automated data reasoning can be achieved. Moreover, the multi-agent reasoning system automatically performs the corresponding data reasoning operation based on the input reasoning problem and reasoning reference data, without the need to specify the reasoning rules, and the data reasoning is more flexible, which can meet various data reasoning requirements.

[0085] In addition, based on the wisdom of each agent, each agent can focus on the processing of different reasoning stages, so as to ensure the reasoning accuracy of each reasoning link and the accuracy of the entire reasoning process and reasoning result. The reasoning process can also be included in the reasoning result, so that the data reasoning conclusion has stronger explainability, which can enable the user to more intuitively and efficiently understand the origin of the reasoning conclusion, and improve the information acquisition efficiency of the user.

[0086] In another embodiment, referring to FIG. 2, after generating the reasoning result, the third agent further includes the step S204 of evaluating the reasoning result to obtain an evaluation conclusion.

[0087] Specifically, the third agent evaluates the reasoning result, specifically, whether the reasoning result obtained by integrating the reasoning path based on the reasoning path sent by the second agent is correct (step S205) in combination with the reasoning problem, the reasoning path and the reasoning key information. In the evaluation conclusion, the judgment information of whether the reasoning result is correct is included, and the explanation information of the incorrect reasoning result can also be included.

[0088] Then, the third agent further executes subsequent operations according to the evaluation conclusion.

[0089] If the evaluation conclusion indicates that the obtained reasoning result is correct, the third agent outputs the reasoning result (step S206).

[0090] If the evaluation conclusion indicates that the obtained reasoning result is incorrect, the third agent sends the evaluation conclusion to the second agent to call the second agent to regenerate the reasoning path (step S102). Correspondingly, the second agent regenerates the reasoning path again according to the reasoning problem and the reasoning key information according to the evaluation conclusion sent by the third agent, and sends the reasoning path to the third agent.

[0091] The third agent integrates the reasoning according to the reasoning problem and the reasoning path regenerated by the second agent, and regenerates the reasoning result corresponding to the reasoning problem.

[0092] Then, the third agent evaluates the regenerated reasoning result again to obtain an evaluation conclusion.

[0093] The above process is repeated until the third agent outputs the generated reasoning result when it judges that the reasoning result is correct, or outputs a reasoning abnormality prompt information when the number of repetitions reaches a preset number.

[0094] Through the above evaluation operation, the third agent further verifies the correctness of the reasoning conclusion, thereby ensuring that a correct reasoning result is generated and improving the accuracy of data reasoning.

[0095] In some other embodiments, the multi-agent reasoning system described above further includes a fourth agent and a fifth agent. The fourth agent is configured to retrieve knowledge required for data reasoning from a knowledge base during data reasoning, and the fifth agent is configured to build and maintain a retrieval-friendly knowledge base for the fourth agent to perform knowledge retrieval.

[0096] The fifth agent builds the knowledge base by performing knowledge segmentation and integration on knowledge source files, and then builds a retrieval-friendly knowledge base. The knowledge source files described above can be any type of knowledge carrier in any field, including but not limited to books, articles, news manuscripts, and network resource files. Specifically, the fifth agent can select knowledge source files in a corresponding field to build a knowledge base according to the field of the built knowledge base or according to the application field of the multi-agent reasoning system.

[0097] In some embodiments, the fifth agent builds the knowledge base by performing the following steps A1-A3:

[0098] A1, performing knowledge segmentation on the knowledge source files to obtain knowledge tags.

[0099] For example, the fifth agent can perform knowledge segment segmentation on the content in the knowledge source files according to chapters, titles, etc. of the content, and perform information summarization or information summarization on each segmented knowledge segment to obtain at least one knowledge tag.

[0100] Alternatively, the fifth agent can perform semantic recognition and classification on the content in the knowledge source files, extract different semantic segments therefrom, and perform information summarization or information summarization on each semantic segment to obtain at least one knowledge tag.

[0101] For example, assuming that the fifth agent builds a medical diagnosis knowledge base, the knowledge base can be built based on a medical textbook. The fifth agent divides the textbook into chapters and paragraphs, and extracts knowledge summaries from each paragraph to obtain corresponding knowledge tags.

[0102] For example, for the section of "coronary atherosclerotic heart disease" in textbook A, the fifth intelligent agent divides it into 3 segments and generates three knowledge tags of "pathophysiology of coronary atherosclerotic heart disease", "clinical manifestations of coronary atherosclerotic heart disease", and "treatment of coronary atherosclerotic heart disease" respectively.

[0103] The above-mentioned section segmentation processing is performed on textbook B, and the knowledge tag of "prevention of coronary atherosclerotic heart disease" can be generated for the segmentation result.

[0104] Similarly, for other medical textbooks and other types of medical knowledge source files, such as medical papers, medical cases, etc., knowledge tags can be generated in the above-mentioned manner.

[0105] A2, generating knowledge triples based on knowledge tags.

[0106] Among them, the knowledge triples include knowledge entities, attribute information of knowledge entities, and sources of knowledge entities and attribute information of knowledge entities.

[0107] Specifically, after obtaining the knowledge tags through the processing of step A1, the fifth intelligent agent respectively processes knowledge collation for each knowledge tag, determines the knowledge entities in the knowledge tags, the attribute information of the knowledge entities, and the sources of the knowledge entities and the attribute information of the knowledge entities, and generates knowledge triples based on the above-mentioned knowledge entities, attribute information, and sources.

[0108] For example, for the four knowledge tags of "pathophysiology of coronary atherosclerotic heart disease", "clinical manifestations of coronary atherosclerotic heart disease", "treatment of coronary atherosclerotic heart disease", and "prevention of coronary atherosclerotic heart disease", the fifth intelligent agent respectively generates corresponding knowledge triples as follows: <coronary atherosclerotic heart disease, pathophysiology, textbook A>, <coronary atherosclerotic heart disease, clinical manifestations, textbook A>, <coronary atherosclerotic heart disease, treatment, textbook A>, <coronary atherosclerotic heart disease, prevention, textbook B>.

[0109] In other embodiments, the attribute information of the knowledge entity includes the chapter content coding of the knowledge entity and the attribute information of the knowledge entity in the source file of the knowledge entity and the knowledge entity, that is, the chapter content of the source file containing the knowledge entity and the attribute information of the knowledge entity is coded, such as generating its md5 code, and linking it as part of the attribute information of the knowledge entity to the knowledge entity. Through the coding, when the attribute information of any knowledge entity is queried, the complete knowledge content containing the knowledge entity and its attribute information can be obtained through the corresponding content coding.

[0110] A3, performing knowledge entity-centered integration processing on each knowledge triple to obtain a knowledge graph.

[0111] Specifically, after obtaining a series of knowledge triples through the processing of step A2, the fifth agent performs knowledge entity-centered integration processing on each knowledge triple to obtain a knowledge graph, so that the attribute information and source information of the same knowledge entity are aggregated together, thereby enabling fast and comprehensive knowledge retrieval based on the constructed knowledge graph.

[0112] For example, for the knowledge entity "coronary atherosclerotic heart disease", the knowledge triples <coronary atherosclerotic heart disease, pathophysiology, textbook A>, <coronary atherosclerotic heart disease, clinical manifestations, textbook A>, <coronary atherosclerotic heart disease, treatment, textbook A>, and <coronary atherosclerotic heart disease, prevention, textbook B> are integrated centered on the knowledge entity "coronary atherosclerotic heart disease", thereby obtaining a knowledge graph centered on the knowledge entity "coronary atherosclerotic heart disease" and having the attribute information and source of the knowledge entity as nodes.

[0113] After constructing the knowledge base, the fifth agent also continuously maintains the knowledge base, such as periodically or in real time updating and supplementing the knowledge base, so that the knowledge of the knowledge base becomes more and more rich, thereby providing a more and more solid knowledge foundation for data reasoning.

[0114] Based on the above-mentioned knowledge base and the above-mentioned multi-agent reasoning system architecture, in the process of generating a reasoning path by the second agent according to a reasoning problem and reasoning key information, a retrieval problem can also be generated, which is used to retrieve the knowledge required for generating and / or executing the reasoning path from the knowledge base.

[0115] Specifically, in the process of determining the reasoning path by the second agent based on the reasoning problem and analyzing the reasoning key information, the second agent further analyzes the knowledge required for generating or executing the reasoning path, and generates a retrieval problem for retrieving the knowledge from the knowledge base.

[0116] In some embodiments, referring to FIG. 3, the second agent determines the reasoning path based on the reasoning problem and analyzes the reasoning key information (step S102), and if additional knowledge is required for executing the reasoning path, the second agent generates a retrieval problem for retrieving the knowledge while generating the reasoning path (step S303), and sends the generated reasoning path to the third agent and the generated retrieval problem to the fourth agent.

[0117] After receiving the retrieval question, the fourth agent retrieves corresponding knowledge from the knowledge base 304 constructed by the fifth agent according to the retrieval question (step S305), and obtains a retrieval result corresponding to the retrieval question.

[0118] In the above retrieval result, relevant knowledge corresponding to the retrieval question is included. In other embodiments, after the fourth agent retrieves corresponding knowledge from the knowledge base constructed by the fifth agent according to the received retrieval question, the fourth agent arranges, induces, and refines the retrieved knowledge data, and takes the arranged, induced, and refined knowledge data as the retrieval result corresponding to the retrieval question.

[0119] In another embodiment, after the fourth agent retrieves relevant knowledge corresponding to the retrieval question from the knowledge base according to the retrieval question, the fourth agent further evaluates the retrieved relevant knowledge, judges the correctness of the relevant knowledge, and obtains evaluation information of the relevant knowledge. Exemplarily, the above evaluation information can be "correct", "incorrect", "ambiguous", etc. The fourth agent sends the evaluation information as part of the retrieval result to the third agent together with the relevant knowledge in the retrieval result.

[0120] After receiving the reasoning question and the reasoning path sent by the second agent, and receiving the retrieval result sent by the fourth agent, the third agent integrates reasoning according to the reasoning question, the retrieval result, and the reasoning path, and generates a reasoning result corresponding to the reasoning question (step S306).

[0121] In the integration reasoning process of the third agent, the relevant knowledge is further combined to integrate the reasoning path, and the evaluation information of the relevant knowledge can also be referred to to determine the influence weight of the relevant knowledge on the reasoning process, so that the third agent can perform more accurate and comprehensive integration reasoning, and improve the accuracy and efficiency of the reasoning result.

[0122] In other embodiments, in the process of determining the reasoning path by the second agent based on the reasoning question and the reasoning key information, some relevant knowledge may need to be referred to to continue determining the reasoning path. In this case, the second agent determines the knowledge required for generating the reasoning path in the process of generating the reasoning path, and generates a retrieval question for retrieving the knowledge from the knowledge base.

[0123] After the second agent generates the retrieval question, the retrieval question is sent to the fourth agent, the fourth agent retrieves relevant knowledge corresponding to the retrieval question from the knowledge base constructed by the fifth agent, and feeds back the retrieval result to the second agent.

[0124] Then, the second agent continues to determine the final reasoning path based on the received retrieval result, in combination with the reasoning question, the previous reasoning conclusion, and the description of the reasoning path.

[0125] The search result includes the relevant knowledge corresponding to the search question sent by the second agent, and can also include evaluation information of the relevant knowledge.

[0126] The pre-inference conclusion is an inference conclusion obtained by the second agent when generating the search question. The inference conclusion can be an incomplete inference path. The second agent can continue to infer the incomplete inference path based on the search, until the final complete inference path is obtained.

[0127] The description of the inference path refers to the explanation or description of the process of generating the inference path by inferring the search key information based on the search question. The second agent generates the description of the inference path based on the understanding of the inference question and the inference key information.

[0128] After receiving the search result, the second agent continues to infer based on the search result to generate a subsequent inference path. In the process of generating the subsequent inference path, if the second agent determines that additional relevant knowledge is needed to obtain the inference path, the second agent can again generate a search question and send it to the fourth agent. The fourth agent searches for the relevant knowledge and sends the search result to the second agent. The second agent continues to infer based on the received search result until the final inference path is generated.

[0129] For example, in a medical diagnosis reasoning scenario, the patient profile information provided for the first agent is "Chief complaint: acid reflux, heartburn for half a month. Present illness: The patient has had acid reflux and heartburn for half a month, which is more obvious after meals, accompanied by retrosternal pain, no vomiting, no nausea, poor appetite, and came to the hospital today. Past history: The patient has been healthy in the past. Physical examination: clear consciousness, normal lung breath sounds, tenderness (+) under xiphoid process, no rebound pain, no muscle guard, no shifting dullness, normal bowel sounds. Laboratory examination: 2020-03-08 gastroscopy: mild damage to esophageal mucosa, follow-up. Age: 57 years old, gender: male", and the diagnosis and treatment reasoning question sent by the first agent is "What do I have?". In the process of reasoning to generate a reasoning path based on the diagnosis and treatment reasoning question by the second agent, when the intermediate conclusion "First, according to the patient's chief complaint and present illness, the patient has acid reflux and heartburn, especially after meals. Moreover, there is retrosternal pain. This suggests that we should consider upper gastrointestinal diseases." is reached, the second agent believes that the upper gastrointestinal disease causing acid reflux and heartburn needs to be further determined in order to continue reasoning. At this time, the second agent generates a retrieval question "Upper gastrointestinal diseases that can cause acid reflux and heartburn" and sends it to the fourth agent. The fourth agent retrieves relevant knowledge "Acid reflux and heartburn may be symptoms of gastroesophageal reflux disease, chronic gastritis, gastric ulcer, etc." from the medical knowledge base constructed by the fifth agent based on the retrieval question, and feeds back the relevant knowledge to the second agent. The second agent continues to reason based on the knowledge and obtains a more perfect reasoning path "First, according to the patient's chief complaint and present illness, the patient has acid reflux and heartburn, especially after meals. Moreover, there is retrosternal pain. This suggests that we should consider upper gastrointestinal diseases. In particular, the possibility of gastroesophageal reflux disease (GERD) and chronic gastritis. The patient's past medical history seems to have no other obvious medical history, and the gastroscopy examination mentioned in the laboratory examination shows mild damage to the esophageal mucosa".

[0130] At this time, the second agent considers that it is still necessary to further determine the upper gastrointestinal disease that can cause mild damage to the esophageal mucosa in order to continue reasoning. The second agent generates a retrieval question "upper gastrointestinal disease that can cause mild damage to the esophageal mucosa" and sends the retrieval question to the fourth agent. The fourth agent retrieves relevant knowledge "esophageal mucosal damage may be caused by diseases such as reflux esophagitis and esophageal ulcer" from the medical knowledge base constructed by the fifth agent based on the retrieval question, and feeds back the relevant knowledge to the second agent. The second agent combines the knowledge to continue reasoning and obtains a more perfect reasoning path: "First, according to the patient's complaint and present illness history, the patient has acid reflux, heartburn, especially after meals. And there is chest pain. This suggests that we should consider upper gastrointestinal diseases. Especially the possibility of gastroesophageal reflux disease (GERD) and chronic gastritis. The patient's past medical history does not seem to have other obvious medical history, so we can temporarily rule out chronic complications or recurrence related to previous diseases. Physical examination found that the xiphoid process was tender, which may indicate gastric inflammation, esophageal or gastric problems. There is no rebound pain or muscle guard, so we can first rule out acute abdominal symptoms related to membrane irritation signs. The gastroscopy mentioned in the examination shows mild damage to the esophageal mucosa, which suggests that gastroesophageal reflux may cause esophagitis. The damage to the esophageal mucosa is consistent with the patient's symptoms of heartburn."

[0131] In another embodiment, referring to FIG. 4, the fourth agent can further determine whether the retrieval question needs to be split when retrieving relevant knowledge from the knowledge base based on the retrieval question (step S404).

[0132] For example, the fourth agent determines whether the retrieval question needs to be split into multiple retrieval sub-questions based on the complexity of the retrieval question after receiving the retrieval question, and then retrieves corresponding knowledge based on the retrieval sub-questions. For example, if the retrieval question is long or needs to retrieve a lot of information, the fourth agent determines that the retrieval question needs to be split. If the retrieval question is short or only needs to retrieve one item of information, the fourth agent determines that the retrieval question does not need to be split.

[0133] Alternatively, the fourth agent performs a trial retrieval based on the retrieval question after receiving the retrieval question, retrieves relevant knowledge from the knowledge base, and then evaluates the relevant knowledge to determine evaluation information. Based on the evaluation information, the fourth agent can determine whether the retrieval question needs to be split. If the evaluation information indicates that the relevant knowledge obtained by the trial retrieval is incorrect or ambiguous, the fourth agent determines that the retrieval question needs to be split in order to retrieve more correct relevant knowledge from the knowledge base through the split retrieval sub-questions. If the evaluation information indicates that the relevant knowledge obtained by the trial retrieval is correct, the fourth agent can determine that the retrieval question does not need to be split.

[0134] In the case of determining that the retrieval question needs to be split, the fourth agent sends a split request to the second agent so as to invoke the second agent to split the retrieval question into multiple retrieval sub-questions.

[0135] After receiving the split request, the second agent splits the retrieval question into sub-questions to obtain multiple sub-questions (step S405), and then sends each sub-question obtained by splitting to the fourth agent. For example, the second agent can parse retrieval intents in the retrieval question through semantic recognition, and split each retrieval intent into a retrieval sub-question.

[0136] The fourth agent retrieves retrieval results corresponding to each sub-question from the knowledge base according to the multiple sub-questions received, and feeds back to the second agent.

[0137] Further, the fourth agent can also integrate and induce the retrieval results corresponding to each sub-question to obtain comprehensive retrieval results and feed back to the second agent.

[0138] For example, in the medical diagnosis reasoning scenario, the retrieval question sent by the second agent to the fourth agent is “Clinical significance of 1, left ventricular wall myocardial segmental motion disorder 2, left heart enlargement, left ventricular systolic dysfunction 3, aortic valve regurgitation (mild), mitral valve regurgitation (mild to moderate) after pacemaker implantation”. After receiving the retrieval question, the fourth agent determines that it needs to be split, and sends a split request to the second agent. After receiving the split request, the second agent splits the above retrieval question into three retrieval sub-questions, which are “Clinical significance of left ventricular wall myocardial segmental motion disorder”, “Clinical significance of left heart enlargement, left ventricular systolic dysfunction”, and “Clinical significance of aortic valve regurgitation (mild), mitral valve regurgitation (mild to moderate)”. Then, the second agent sends the above three retrieval sub-questions to the fourth agent, and the fourth agent retrieves relevant knowledge corresponding to each retrieval sub-question from the knowledge base according to each retrieval sub-question. On this basis, the fourth agent integrates the relevant knowledge corresponding to each retrieval sub-question to obtain the final retrieval result “Clinical significance of 1, left ventricular wall myocardial segmental motion disorder 2, left heart enlargement, left ventricular systolic dysfunction 3, aortic valve regurgitation (mild), mitral valve regurgitation (mild to moderate) after pacemaker implantation is heart function reduction and valve regurgitation”.

[0139] Splitting the retrieval question into sub-questions for retrieval can improve the coverage and accuracy of knowledge retrieval, and thus can improve the accuracy of data reasoning.

[0140] Next, the complete processing flow of the data reasoning method performed by the multi-agent reasoning system according to the embodiments of the present application will be introduced in combination with the reasoning process in the medical diagnosis and treatment reasoning scenario.

[0141] Referring to FIG. 5, the fifth agent pre-constructs a medical knowledge base containing various disease knowledge (step S501). When the first agent receives a diagnosis reasoning question and medical documents of a patient, the first agent extracts key information from the medical documents of the patient according to the diagnosis reasoning question to obtain profile information of the patient (step S502), wherein the basic profile information of the patient is protected, and the time sequence profile information of the patient is obtained. Then, the first agent sends the diagnosis reasoning question and the profile information of the patient to the second agent.

[0142] The second agent generates a reasoning path according to the diagnosis reasoning question and the profile information of the patient (step S503), and determines whether reference knowledge needs to be searched. If the reference knowledge needs to be searched, a search question is generated (step S504) and sent to the fourth agent.

[0143] The fourth agent receives the search question and searches the knowledge from the medical knowledge base 505 constructed by the fifth agent, and determines whether the search question needs to be split (step S506). If the search question needs to be split, a splitting instruction is generated and sent to the second agent.

[0144] The second agent receives the splitting instruction and splits the search question into multiple search sub-questions (step S507), and sends each search sub-question obtained by splitting to the fourth agent.

[0145] The fourth agent searches the relevant knowledge from the knowledge base based on each search sub-question (step S508), and generates a search result based on the relevant knowledge obtained by searching. In the search result, the evaluation information of the fourth agent on the relevant knowledge obtained by searching can also be included.

[0146] The reasoning path generated by the second agent and the search result generated by the fourth agent are sent to the third agent together with the diagnosis reasoning question, and the third agent integrates and reasons to obtain a reasoning result (step S509) and evaluates the reasoning result to obtain an evaluation conclusion (step S510), and then determines whether the reasoning result is correct (step S511).

[0147] If the evaluation conclusion indicates that the reasoning result is correct, the third agent outputs the reasoning result (step S512).

[0148] If the evaluation conclusion indicates that the reasoning result is incorrect, the third agent sends a re-reasoning instruction to the second agent, so that the second agent generates a reasoning path according to the diagnosis reasoning question and the profile information of the patient again. Each agent re-executes the above reasoning process until the third agent considers that the obtained reasoning result is correct, and outputs the reasoning result.

[0149] The above embodiments only introduce the processing flow of the data reasoning method proposed in the embodiments of the present application in the medical diagnosis and treatment scene from a global perspective, so that those skilled in the art can more comprehensively understand the complete processing flow of the data reasoning method proposed in the embodiments of the present application. For the specific processing content of each step in the flow, please refer to the specific processing content of the corresponding processing step in the above embodiments, which will not be repeated here.

[0150] In the above embodiments, by introducing the work content and collaborative processing process of each agent in the multi-agent reasoning system in the data reasoning process, the data reasoning method proposed in the present application is exemplarily introduced. Based on the above embodiment introduction, those skilled in the art can know the composition structure of the multi-agent reasoning system proposed in the present application and the specific functions of each part of the structure.

[0151] For example, the embodiments of the present application propose a multi-agent reasoning system, which includes a plurality of agents. The plurality of agents in the multi-agent reasoning system are configured to implement the data reasoning method introduced in any of the above embodiments.

[0152] In some embodiments, the multi-agent reasoning system includes a first agent, a second agent and a third agent.

[0153] The first agent is configured to extract key information from the reasoning reference data according to the reasoning problem, to obtain reasoning key information; the second agent is configured to generate a reasoning path according to the reasoning problem and the reasoning key information; the reasoning path includes prompt information for reasoning based on the reasoning key information to answer the reasoning problem; and the third agent is configured to integrate reasoning according to the reasoning problem and the reasoning path, to generate a reasoning result corresponding to the reasoning problem, wherein the reasoning result includes a reasoning process and / or a reasoning conclusion.

[0154] In another embodiment, the third agent is further configured to evaluate the reasoning result to obtain an evaluation conclusion; in the case that the evaluation conclusion represents that the reasoning result is wrong, the third agent calls the second agent to regenerate the reasoning path, and the third agent integrates reasoning according to the reasoning problem and the regenerated reasoning path of the second agent, to regenerate the reasoning result corresponding to the reasoning problem.

[0155] In another embodiment, the multi-agent reasoning system further includes a fourth agent and a fifth agent; during the process of generating a reasoning path based on the reasoning question and key reasoning information, the second agent also generates a retrieval question, which is used to retrieve the knowledge required to execute the reasoning path from the knowledge base; the fourth agent is used to retrieve the retrieval result corresponding to the retrieval question from the knowledge base based on the retrieval question; the retrieval result includes relevant knowledge corresponding to the retrieval question, and the knowledge base is constructed by the fifth agent; the third agent performs integrated reasoning based on the reasoning question and the reasoning path to generate the reasoning result corresponding to the reasoning question, including: the third agent performs integrated reasoning based on the reasoning question, the retrieval result, and the reasoning path to generate the reasoning result corresponding to the reasoning question.

[0156] In another embodiment, the fifth agent constructs a knowledge base by performing the following processes: segmenting the knowledge source files to obtain knowledge tags; generating knowledge triples based on the knowledge tags, wherein each knowledge triple includes a knowledge entity, attribute information of the knowledge entity, and the source of the knowledge entity and its attribute information; and performing an integration process centered on the knowledge entity on each knowledge triple to obtain a knowledge graph.

[0157] In another embodiment, the attribute information of the knowledge entity includes the encoding of the knowledge entity and the attribute information of the knowledge entity in the chapter content of the knowledge entity and the source file of the knowledge entity.

[0158] In another embodiment, the search results also include evaluation information on the relevant knowledge.

[0159] In another embodiment, the fourth agent is further configured to determine whether the retrieval question needs to be split; if it is determined that the retrieval question needs to be split, the fourth agent calls the second agent to split the retrieval question into multiple retrieval sub-questions; and the fourth agent retrieves retrieval results corresponding to each retrieval sub-question from the knowledge base based on the multiple retrieval sub-questions.

[0160] In another embodiment, the reasoning problem includes a diagnostic reasoning problem, and the reasoning reference data includes the patient's medical records.

[0161] In another embodiment, the patient's medical documents include multiple medical documents from different periods; the first intelligent agent extracts key information from the reasoning reference data according to the reasoning problem to obtain key reasoning information, including: the first intelligent agent extracts patient profile information from multiple medical documents according to the reasoning problem to generate a basic information profile and a time-series medical profile of the corresponding patient.

[0162] For the specific processing procedures of each intelligent agent in the above embodiments, as well as the implementation form and training method of each intelligent agent, please refer to the above-described embodiments of the data reasoning method, which will not be repeated here.

[0163] Corresponding to the above-described data reasoning method, another embodiment of this application proposes a data reasoning device. The device is applied to a multi-agent reasoning system comprising multiple agents. The device includes: an information integration unit, used by a first agent to extract key information from reasoning reference data based on the reasoning problem, obtaining key reasoning information; an analysis and reasoning unit, used by a second agent to generate a reasoning path based on the reasoning problem and the key reasoning information; the reasoning path includes prompts for reasoning to answer the reasoning problem based on the key reasoning information; and an integrated reasoning unit, used by a third agent to perform integrated reasoning based on the reasoning problem and the reasoning path, generating a reasoning result corresponding to the reasoning problem, the reasoning result including the reasoning process and / or the reasoning conclusion.

[0164] In some embodiments, the integrated reasoning unit is further configured to: evaluate the reasoning result by a third agent to obtain an evaluation conclusion; if the evaluation conclusion indicates that the reasoning result is incorrect, call the analysis reasoning unit through the third agent to regenerate the reasoning path through the second agent; and the integrated reasoning unit performs integrated reasoning by the third agent based on the reasoning problem and the reasoning path regenerated by the second agent to regenerate the reasoning result for the corresponding reasoning problem.

[0165] In some embodiments, during the process of generating a reasoning path based on a reasoning question and key reasoning information by a second intelligent agent, the analysis and reasoning unit also generates a retrieval question. The retrieval question is used to retrieve the knowledge required to execute the reasoning path from a knowledge base. The apparatus further includes: a knowledge retrieval unit, used by a fourth intelligent agent to retrieve retrieval results corresponding to the retrieval question from the knowledge base; the retrieval results include relevant knowledge corresponding to the retrieval question, and the knowledge base is constructed by a fifth intelligent agent; the integration and reasoning unit performs integration and reasoning based on the reasoning question and the reasoning path by a third intelligent agent to generate a reasoning result corresponding to the reasoning question, including: performing integration and reasoning based on the reasoning question, the retrieval result, and the reasoning path by the third intelligent agent to generate a reasoning result corresponding to the reasoning question.

[0166] In some embodiments, the fifth agent constructs a knowledge base by performing the following processes: segmenting the knowledge source files to obtain knowledge tags; generating knowledge triples based on the knowledge tags, wherein each knowledge triple includes a knowledge entity, attribute information of the knowledge entity, and the source of the knowledge entity and its attribute information; and performing an integration process centered on the knowledge entity on each knowledge triple to obtain a knowledge graph.

[0167] In some embodiments, the attribute information of a knowledge entity includes the encoding of the knowledge entity and its attribute information in the chapter content of the knowledge entity and its source file.

[0168] In some embodiments, the search results may also include evaluation information on the relevant knowledge.

[0169] In some embodiments, the fourth agent further determines whether the retrieval question needs to be split; if it is determined that the retrieval question needs to be split, the knowledge retrieval unit calls the second agent through the fourth agent to split the retrieval question into multiple retrieval sub-questions; and the knowledge retrieval unit retrieves retrieval results corresponding to each retrieval sub-question from the knowledge base through the fourth agent based on the multiple retrieval sub-questions.

[0170] In some embodiments, the reasoning problem includes a diagnostic reasoning problem, and the reasoning reference data includes the patient's medical records.

[0171] In some embodiments, the patient's medical documents include multiple medical documents from different periods; the information integration unit extracts key information from the reasoning reference data based on the reasoning problem by a first intelligent agent to obtain key reasoning information, including: extracting patient profile information from multiple medical documents based on the reasoning problem by the first intelligent agent to generate a basic information profile and a time-series medical profile of the corresponding patient.

[0172] The data inference apparatus provided in this embodiment belongs to the same concept as the data inference method provided in the above embodiments of this application. It can execute the data inference method provided in any of the above embodiments of this application and has the corresponding functional units and beneficial effects of the execution method. Technical details not described in detail in this embodiment can be found in the specific processing content of the data inference method provided in the above embodiments of this application, and will not be repeated here.

[0173] Another embodiment of this application also proposes a data reasoning device, which includes an input device, an output device, and a multi-agent reasoning system; the input device is connected to a processor and is used to input a reasoning problem and reasoning reference data to the processor; the output device is connected to the processor and is used to output the reasoning result generated by the processor; the multi-agent reasoning system is configured to execute the data reasoning method of any of the above embodiments.

[0174] The aforementioned input devices may include devices for receiving user input data and information, such as keyboards, mice, cameras, scanners, light pens, voice input devices, touch screens, pedometers, or gravity sensors.

[0175] The aforementioned output devices may include devices that allow information to be output to the user, such as displays, printers, speakers, etc.

[0176] For details regarding the specific structure and function of the multi-agent reasoning system described above, as well as its specific processing procedure when executing the data reasoning method of any of the above embodiments, please refer to the corresponding descriptions of the above method embodiments.

[0177] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions. The computer program instructions may include multiple segments or groups of computer program instructions, which are executed by multiple agents in a multi-agent inference system to cause the multi-agent inference system to execute the data inference method of any of the above embodiments.

[0178] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0179] Furthermore, embodiments of this application may also be storage media storing computer programs. The computer programs may include multiple segments or groups of computer programs. When these computer programs are run by multiple agents in a multi-agent inference system, the multi-agent inference system executes the data inference method described in any of the above embodiments.

[0180] Another embodiment of this application proposes another data reasoning method. The data reasoning process of this method is the same as that of the data reasoning method described in the above embodiments. The difference is that the data reasoning method proposed in this application embodiment can be executed by a single intelligent agent or on a single device, such as on a processor, computer, server, smart terminal or other device or equipment.

[0181] For example, the single agent or single device used to implement the data reasoning method proposed in the embodiments of this application can be an agent integrated from multiple agents in the multi-agent reasoning system, or a device that runs multiple agents simultaneously, such as a server or workstation. Alternatively, multiple data reasoning models can be deployed on a single agent or device, each model corresponding to a specific agent in the multi-agent reasoning system, and used to implement the functions of each agent in the multi-agent reasoning system and execute the processing of each agent.

[0182] Referring to Figure 6, the data reasoning method proposed in this application includes:

[0183] S601. Extract key information from the reasoning reference data based on the reasoning problem to obtain key reasoning information.

[0184] S602. Generate a reasoning path based on the reasoning question and key reasoning information. The reasoning path includes hints that use the key reasoning information to reason and solve the reasoning question.

[0185] S603. Based on the reasoning problem and reasoning path, perform integrated reasoning to generate the reasoning result corresponding to the reasoning problem. The reasoning result includes the reasoning process and / or reasoning conclusion.

[0186] In another embodiment, the method further includes: evaluating the reasoning result to obtain an evaluation conclusion; if the evaluation conclusion indicates that the reasoning result is incorrect, regenerating the reasoning path based on the reasoning question and key reasoning information; and performing integrated reasoning based on the reasoning question and the regenerated reasoning path to regenerate the reasoning result corresponding to the reasoning question.

[0187] In another embodiment, the method further includes: generating a retrieval question based on a reasoning question and key reasoning information, the retrieval question being used to retrieve the knowledge required to execute the reasoning path from a knowledge base; retrieving retrieval results corresponding to the retrieval question from the knowledge base based on the retrieval question; the retrieval results including relevant knowledge corresponding to the retrieval question; and performing integrated reasoning based on the reasoning question and the reasoning path to generate a reasoning result corresponding to the reasoning question, including: performing integrated reasoning based on the reasoning question, the retrieval results, and the reasoning path to generate a reasoning result corresponding to the reasoning question.

[0188] In another embodiment, the search results also include evaluation information on the relevant knowledge.

[0189] In another embodiment, the method further includes: determining whether the retrieval question needs to be split; if it is determined that the retrieval question needs to be split, splitting the retrieval question into multiple retrieval sub-questions; and retrieving retrieval results corresponding to each retrieval sub-question from the knowledge base based on the multiple retrieval sub-questions.

[0190] In another embodiment, the reasoning problem includes a diagnostic reasoning problem, and the reasoning reference data includes the patient's medical records.

[0191] In another embodiment, the patient's medical documents include multiple medical documents from different periods; key information is extracted from the reasoning reference data based on the reasoning question to obtain key reasoning information, including: extracting patient profile information from multiple medical documents based on the reasoning question to generate a basic information profile and a time-series medical profile of the corresponding patient.

[0192] The data reasoning method described in the above embodiments belongs to the same application concept as the data reasoning method applied to a multi-agent reasoning system. For the specific processing procedures of each processing step, please refer to the specific processing procedures of the corresponding processing steps in the data reasoning method applied to a multi-agent reasoning system.

[0193] By executing the aforementioned data reasoning method through a single intelligent agent or device, the flexibility and efficiency of data reasoning can be improved. Users can provide any reasoning problem and reference data, enabling the intelligent agent or device to execute the data reasoning method, realize the data reasoning process, and obtain the reasoning result corresponding to the reasoning problem. The entire reasoning process does not require specifying reasoning rules, making data reasoning more flexible and able to meet various different data reasoning needs. The reasoning result can also include the reasoning process, thereby making the data reasoning conclusion more interpretable, allowing users to understand the origin of the reasoning conclusion more intuitively and efficiently, and improving the user's information acquisition efficiency.

[0194] On the other hand, this method can be executed by a single intelligent agent or device, which reduces the implementation cost of the method.

[0195] Corresponding to the above-described data reasoning method, this application also proposes a data reasoning device, as shown in Figure 7. The device includes: an information integration unit 700, used to extract key information from reasoning reference data according to the reasoning problem to obtain key reasoning information; an analysis and reasoning unit 710, used to generate a reasoning path according to the reasoning problem and the key reasoning information; the reasoning path includes prompts for reasoning and answering the reasoning problem based on the key reasoning information; and an integration and reasoning unit 720, used to perform integration and reasoning according to the reasoning problem and the reasoning path to generate a reasoning result corresponding to the reasoning problem, the reasoning result including the reasoning process and / or the reasoning conclusion.

[0196] In some implementations, the integrated reasoning unit 720 is also used to evaluate the reasoning results and obtain an evaluation conclusion.

[0197] If the evaluation conclusion represents an error in the reasoning result, the integration reasoning unit 720 calls the analysis reasoning unit 710 to regenerate the reasoning path based on the reasoning question and key reasoning information; and the integration reasoning unit 720 performs integration reasoning based on the reasoning question and the regenerated reasoning path to regenerate the reasoning result for the corresponding reasoning question.

[0198] In some implementations, the analysis and reasoning unit 710 is further used to generate a retrieval question based on the reasoning question and key reasoning information. The retrieval question is used to retrieve the knowledge required to execute the reasoning path from the knowledge base. The device also includes a knowledge retrieval unit, used to retrieve retrieval results corresponding to the retrieval question from the knowledge base based on the retrieval question. The retrieval results include relevant knowledge corresponding to the retrieval question. The integration and reasoning unit 720 performs integration and reasoning based on the reasoning question and the reasoning path to generate a reasoning result corresponding to the reasoning question, including: performing integration and reasoning based on the reasoning question, the retrieval result, and the reasoning path to generate a reasoning result corresponding to the reasoning question.

[0199] In some implementations, the search results also include evaluation information on the relevant knowledge.

[0200] In some implementations, the knowledge retrieval unit is also used to determine whether the retrieval question needs to be split; if it is determined that the retrieval question needs to be split, the knowledge retrieval unit calls the analysis and reasoning unit 710 to split the retrieval question into multiple retrieval sub-questions; the knowledge retrieval unit retrieves the retrieval results corresponding to each retrieval sub-question from the knowledge base based on the multiple retrieval sub-questions.

[0201] In some implementations, the reasoning problem includes diagnostic reasoning, and the reasoning reference data includes the patient's medical records.

[0202] In some implementations, the patient's medical documents include multiple medical documents from different periods; the information integration unit 700 extracts key information from the reasoning reference data based on the reasoning problem to obtain key reasoning information, including: extracting patient profile information from multiple medical documents based on the reasoning problem to generate a basic information profile and a time-series medical profile of the corresponding patient.

[0203] The data inference apparatus provided in this embodiment belongs to the same concept as the data inference method provided in the above embodiments of this application. It can execute the data inference method provided in any of the above embodiments of this application and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment can be found in the specific processing content of the data inference method provided in the above embodiments of this application, and will not be repeated here.

[0204] The functions implemented by each of the above units can be implemented by the same or different processors, and this application embodiment does not limit this.

[0205] It should be understood that the units in the above device can be implemented by a processor calling software. For example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of each unit in the device. The processor can be a general-purpose processor, such as a CPU or microprocessor, and the memory can be internal or external to the device. Alternatively, the units in the device can be implemented as hardware circuits. By designing the hardware circuits, some or all of the unit functions can be implemented. The hardware circuits can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above units are implemented by designing the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a PLD, such as an FPGA, which can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files to implement the functions of some or all of the above units. All units in the above device can be implemented entirely by a processor calling software, entirely by hardware circuits, or partially by a processor calling software with the remaining parts implemented by hardware circuits.

[0206] In this application embodiment, a processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction reading and execution capabilities, such as a CPU, microprocessor, GPU, or DSP. In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. These logical relationships are fixed or reconfigurable. For example, the processor may be a hardware circuit implemented as an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the processor loading instructions to implement the functions of some or all of the above units. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, or DPU.

[0207] As can be seen, each unit in the above device can be one or more processors (or processing circuits) configured to implement the above methods, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0208] Furthermore, the units in the above devices can be integrated in whole or in part, or they can be implemented independently. In one implementation, these units are integrated together and implemented in the form of a System-on-Chip (SoC). The SoC may include at least one processor for implementing any of the above methods or implementing the functions of the units in the device. The at least one processor may be of different types, such as CPU and FPGA, CPU and artificial intelligence processor, CPU and GPU, etc.

[0209] Another embodiment of this application also proposes a data reasoning device, as shown in FIG8. The device includes a memory 800 and a processor 810. The memory 800 is connected to the processor 810 and is used to store programs. The processor 810 is used to implement the data reasoning method disclosed in any of the above embodiments by running the programs stored in the memory 800.

[0210] Specifically, the aforementioned data inference device may also include: a bus, a communication interface 820, an input device 830, and an output device 840.

[0211] The processor 810, memory 800, communication interface 820, input device 830, and output device 840 are interconnected via a bus. The bus may include a pathway for transmitting information between various components of the computer system.

[0212] Processor 810 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Processor 810 may include a main processor, and may also include a baseband chip, a modem, etc.

[0213] The memory 800 stores a program that executes the technical solution of this invention, and may also store an operating system and other key business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory 800 may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0214] Input device 830 may include a device for receiving user input data and information, such as a keyboard, mouse, camera, scanner, light pen, voice input device, touch screen, pedometer, or gravity sensor.

[0215] Output device 840 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0216] The communication interface 820 may include a device that uses any transceiver to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0217] The processor 810 executes the program stored in the memory 800 and calls other devices, which can be used to implement the various steps of any of the data reasoning methods provided in the above embodiments of this application.

[0218] This application also proposes a chip, which includes a processor and a data interface. The processor reads and runs a program stored in the memory through the data interface to execute the data reasoning method described in any of the above embodiments. For details of the processing and its beneficial effects, please refer to the embodiments of the above data reasoning method.

[0219] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the data reasoning methods described in any of the above embodiments of this specification.

[0220] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0221] Furthermore, embodiments of this application may also be storage media storing a computer program, which is executed by a processor through steps in the data reasoning method described in any of the above embodiments of this specification.

[0222] For the foregoing method embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0223] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0224] The steps in the methods of the various embodiments of this application can be adjusted, merged, or deleted in order according to actual needs, and the technical features described in each embodiment can be replaced or combined.

[0225] The modules and sub-modules in the various embodiments of the present application's devices and terminals can be merged, divided, and deleted according to actual needs.

[0226] It should be understood that the disclosed terminals, devices, and methods can be implemented in other ways, given the several embodiments provided in this application. For example, the terminal embodiments described above are merely illustrative. For instance, the division of modules or sub-modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple sub-modules or modules may be combined or integrated into another module, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0227] The modules or submodules described as separate components may or may not be physically separate. The components that constitute a module or submodule may or may not be physical modules or submodules; that is, they may be located in one place or distributed across multiple network modules or submodules. Some or all of the modules or submodules can be selected to achieve the purpose of this embodiment's solution, depending on actual needs.

[0228] Furthermore, the functional modules or sub-modules in the various embodiments of this application can be integrated into one processing module, or each module or sub-module can exist physically separately, or two or more modules or sub-modules can be integrated into one module. The integrated modules or sub-modules described above can be implemented in hardware or in the form of software functional modules or sub-modules.

[0229] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0230] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software unit executed by a processor, or a combination of both. The software unit can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0231] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0232] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data inference method, characterized by, Applied to a multi-agent reasoning system comprising a plurality of agents, the method comprises: a first agent extracts key information from reasoning reference data according to a reasoning question, to obtain reasoning key information; a second agent generates a reasoning path according to the reasoning question and the reasoning key information; the reasoning path comprises prompt information for reasoning based on the reasoning key information to answer the reasoning question; a third agent integrates reasoning according to the reasoning question and the reasoning path, to generate a reasoning result corresponding to the reasoning question, the reasoning result comprising a reasoning process and / or a reasoning conclusion.

2. The method of claim 1, wherein, The method further comprises: the third agent evaluates the reasoning result to obtain an evaluation conclusion; in a case where the evaluation conclusion represents that the reasoning result is incorrect, the third agent calls the second agent to re-generate a reasoning path, and the third agent integrates reasoning according to the reasoning question and the re-generated reasoning path of the second agent, to re-generate a reasoning result corresponding to the reasoning question.

3. The method of claim 1, wherein, In the process of generating a reasoning path by the second agent according to the reasoning question and the reasoning key information, a retrieval question is further generated, the retrieval question being used to retrieve knowledge required for executing the reasoning path from a knowledge base.

4. The method of claim 3, wherein, The method further comprises: a fourth agent retrieves a retrieval result corresponding to the retrieval question from the knowledge base according to the retrieval question; the retrieval result comprises relevant knowledge corresponding to the retrieval question, and the knowledge base is constructed by a fifth agent.

5. The method of claim 4, wherein, The third agent integrates reasoning according to the reasoning question, the retrieval result, and the reasoning path, to generate a reasoning result corresponding to the reasoning question. The fifth agent constructs the knowledge base by performing the following processing:

6. The method of claim 4, wherein, performing knowledge segmentation on a knowledge source file to obtain a knowledge label; generating a knowledge triple based on the knowledge label, the knowledge triple comprising a knowledge entity, attribute information of the knowledge entity, and a source of the knowledge entity and the attribute information of the knowledge entity; performing integration processing on each knowledge triple with the knowledge entity as the center to obtain a knowledge graph. The attribute information of the knowledge entity comprises an encoding of chapter content in which the knowledge entity and the attribute information of the knowledge entity are located in a source file of the knowledge entity and the attribute information of the knowledge entity.

7. The method of claim 6, wherein, The retrieval result further comprises evaluation information of the relevant knowledge.

8. The method according to any one of claims 4 to 7, characterized in that, The fourth agent retrieves a retrieval result corresponding to the retrieval question from the knowledge base according to the retrieval problem, comprising:

9. The method according to any one of claims 4 to 8, characterized in that, the fourth agent determines whether the retrieval question needs to be split; in a case where it is determined that the retrieval question needs to be split, the fourth agent calls the second agent to split the retrieval question into a plurality of retrieval sub-questions; and the fourth agent retrieves a retrieval result corresponding to each retrieval sub-question from the knowledge base according to the plurality of retrieval sub-questions. ​ 10. The method according to any one of claims 1 to 9, characterized in that, The reasoning problem includes a diagnosis and treatment reasoning problem, and the reasoning reference data includes medical documents of a patient.

11. The method of claim 10, wherein, The medical documents of the patient include multiple medical documents of the patient at different time periods. The first agent extracts key information from the reasoning reference data according to the reasoning problem to obtain reasoning key information, including: The first agent extracts patient portrait information from the multiple medical documents according to the reasoning problem to generate a basic information portrait and a time sequence medical portrait corresponding to the patient.

12. A multi-agent reasoning system, comprising: The multi-agent system includes a plurality of agents, and the plurality of agents in the multi-agent system are configured to implement the data reasoning method according to any one of claims 1-11.

13. A data inference apparatus, comprising: The device is applied to a multi-agent reasoning system including a plurality of agents, and the device includes: An information integration unit configured to extract key information from the reasoning reference data according to the reasoning problem by a first agent to obtain reasoning key information; An analysis reasoning unit configured to generate a reasoning path according to the reasoning problem and the reasoning key information by a second agent; the reasoning path includes prompt information for answering the reasoning problem based on the reasoning key information; An integrated reasoning unit configured to perform integrated reasoning according to the reasoning problem and the reasoning path by a third agent to generate a reasoning result corresponding to the reasoning problem, the reasoning result including a reasoning process and / or a reasoning conclusion.

14. A data inference device, comprising: The device includes: An input device, an output device, and a multi-agent reasoning system; The input device is connected to the processor and is configured to input a reasoning problem and reasoning reference data to the processor; The output device is connected to the processor and is configured to output a reasoning result generated by the processor; The multi-agent reasoning system is configured to perform the data reasoning method according to any one of claims 1-11.

15. A computer program product, characterised in that, The computer program instructions, when executed by a plurality of agents in a multi-agent reasoning system, cause the multi-agent reasoning system to perform the data reasoning method according to any one of claims 1-11.

Images