Medical information query method based on hint engineering and storage medium

Abstract

The application relates to a medical information query method based on a prompt engineering and a storage medium, and the method comprises the following steps: in response to a query operation triggered by a target object on a query system, obtaining query request information input by the target object based on a prompt engineering, wherein the query request information comprises intention demand information and query category label information; based on the query category label information, at least using a natural language processing (NLP) algorithm to obtain candidate medical information associated with the intention demand information in a target database; in a preset matching rule, a target matching rule corresponding to each query operation is determined, and according to the target matching rule, target medical information corresponding to the intention demand information is screened out from the candidate medical information, and the target medical information is taken as a query result. Through the application, the problem that the efficiency and accuracy of obtaining medical information are low in the scheme for providing medical information by using a large language model in the related art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and medical cross, in particular to a medical information query method based on prompt engineering and a storage medium. BACKGROUND

[0002] With the development of medical informatization, the growth rate of medical data far exceeds the learning and absorption capacity of doctors and medical students. Traditional medical information acquisition methods, including textbooks, medical literature and expert consultation, often have problems such as information update not timely, long query time, uneven resource distribution, etc. Especially in resource-limited environments such as remote areas or emergency situations, medical information cannot be conveniently obtained.

[0003] In related technologies, although a large language model (LLM) can quickly and accurately process and generate a large amount of medical information; however, in related technologies, the scheme of providing medical information by using a large language model cannot meet the diversified acquisition needs of different user groups for medical information, and the efficiency and accuracy of acquiring medical information are low.

[0004] For the problem of low efficiency and accuracy of acquiring medical information in the scheme of providing medical information by using a large language model in related technologies, an effective solution has not been proposed. SUMMARY

[0005] The embodiments of the present application provide a medical information query method based on prompt engineering and a storage medium, and an apparatus and an electronic device, to at least solve the problem of low efficiency and accuracy of acquiring medical information in the scheme of providing medical information by using a large language model in related technologies.

[0006] In a first aspect, the embodiments of the present application provide a medical information query method based on prompt engineering, comprising: in response to a query operation triggered by a target object on a query system, acquiring query request information input by the target object based on prompt engineering, wherein the query request information includes intention demand information and query category label information, and the query category label information is used to represent one of the following query operations: case screening, medical popularization, information diagnosis and consultation interaction; based on the query category label information, at least using a natural language processing (NLP) algorithm to acquire candidate medical information associated with the intention demand information in a target database; in a preset matching rule, determining a target matching rule corresponding to each query operation, and according to the target matching rule, screening target medical information corresponding to the intention demand information from the candidate medical information, and taking the target medical information as a query result.

[0007] In a second aspect, the embodiments of the present application provide a storage medium having a computer program stored thereon, the program being executed by a processor to implement the steps of the medical information query method based on prompt engineering according to the first aspect.

[0008] Compared with the related art, the medical information query method based on prompt engineering and the storage medium provided by the embodiments of the present application have the following advantages. After a target object responds to a query operation triggered by a query system, the query request information input by the target object is obtained based on prompt engineering, wherein the query request information includes intention demand information and query category label information. Based on the query category label information, candidate medical information associated with the intention demand information is obtained in a target database at least by using a natural language processing (NLP) algorithm. In a preset matching rule, a target matching rule corresponding to each query operation is determined, and target medical information corresponding to the intention demand information is filtered from the candidate medical information according to the target matching rule, and the target medical information is taken as a query result. The scheme of providing medical information by using a large language model in the related art is solved, and the efficiency and accuracy of obtaining medical information are improved. Personalized medical knowledge and suggestions are provided for different user groups, the efficiency and accuracy of information acquisition are improved, medical information can be conveniently obtained in a resource-limited environment, the accessibility and convenience of medical services are improved, and the diversified needs of users for medical information acquisition are met.

[0009] Details of one or more embodiments of the present application are presented in the following drawings and description to make other features, objects and advantages of the present application more apparent. BRIEF DESCRIPTION OF DRAWINGS

[0010] The accompanying drawings illustrated herein are used to provide further understanding of the present application, constitute a part of the present application, and the illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0011] Figure 1 is a hardware structure block diagram of a terminal of the medical information query method based on prompt engineering according to the embodiments of the present application;

[0012] Figure 2 is a flowchart of the medical information query method based on prompt engineering according to the embodiments of the present application;

[0013] Figure 3 is a structure block diagram of the medical information acquisition device based on prompt engineering according to the embodiments of the present application. DETAILED DESCRIPTION

[0014] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be described and explained below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application and should not be used to limit the present application. Based on the embodiments provided by the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application. In addition, it should be understood that, although the efforts made in this development process can be complex and lengthy, some design, manufacture or production changes made on the basis of the technical content disclosed in the present application by those of ordinary skill in the art related to the content disclosed in the present application are only routine technical means and should not be understood as insufficient disclosure of the present application.

[0015] In the present application, the term "embodiment" means that the specific features, structures or properties described in conjunction with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment to other embodiments. Those of ordinary skill in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments without conflict.

[0016] Unless otherwise defined, the technical terms or scientific terms involved in the present application should be understood as the usual meaning understood by those of ordinary skill in the art to which the present application belongs. The terms "one", "a", "an", "the" and the like similar words involved in the present application do not represent a quantity limitation, but can represent a singular or plural number. The terms "include", "contain", "have" and any variations thereof involved in the present application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device including a series of steps or modules (units) is not limited to the listed steps or units, but can also include steps or units not listed, or can also include other steps or units inherent to these processes, methods, products or devices. The "multiple links" referred to in the present application refers to more than or equal to two links. "And / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can mean that A exists alone, A and B exist together, and B exists alone. The terms "first", "second", "third" and the like referred to in the present application are only to distinguish similar objects, and do not represent a specific order for the objects.

[0017] Before the embodiments of the present application are described, the related technologies involved in the embodiments of the present application are described as follows:

[0018] The method embodiments provided by the present embodiment can be executed in a terminal, a computer or a similar computing device. Taking the case of running on a terminal,Figure 1 is a hardware structure block diagram of a terminal of a medical information query method based on hint engineering according to an embodiment of the present application. As shown in Figure 1 , the terminal can include one or more (only one is shown in Figure 1 ) processors 102 (the processor 102 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the above terminal can further include a transmission device 106 for communication function and an input and output device 108. Those skilled in the art can understand that Figure 1 the structure shown is only schematic, which does not limit the structure of the above terminal. For example, the terminal can further include more or less components than those shown in Figure 1 , or have a different configuration from Figure 1 .

[0019] The memory 104 can be used to store computer programs, for example, software programs of application software and modules, such as a computer program corresponding to the medical information query method based on hint engineering in the embodiment of the present application. The processor 102 performs various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above method. The memory 104 can include a high-speed random access memory, and can further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor 102, which can be connected to the terminal 10 through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0020] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network can include a wireless network provided by a communication provider of the terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC) which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, RF) module which is used to communicate with the Internet in a wireless manner.

[0021] The embodiment provides a medical information query method based on hint engineering running on the above terminal, Figure 2 is a flowchart of a medical information query method based on hint engineering according to an embodiment of the present application. As shown in Figure 2 , the flowchart includes the following steps:

[0022] Step S201, in response to a target object triggering a query operation in a query system, query request information input by the target object is obtained based on a prompting project, wherein the query request information comprises intention demand information and query category label information, and the query category label information is used to represent one of the following query operations: case screening, medical popular science, information diagnosis, and consultation interaction.

[0023] In the present embodiment, the execution subject of the medical information query method of the present application includes but is not limited to a built intelligent medical platform, and is integrated with a case screening module, a medical popular science module, an information diagnosis module (primary diagnosis), and a consultation interaction module; and is respectively used for different target objects; in the present embodiment, the target objects include medical practitioners (the intelligent medical platform provides professional and detailed case analysis and diagnosis suggestions, supports multiple rounds of in-depth discussions, and helps doctors make more accurate judgments in complex cases), medical institutions, medical students (the intelligent medical platform provides learning-oriented case analysis and knowledge point supplementation, in combination with the medical popular science module, to help medical students acquire systematic and hierarchical knowledge during the learning process), patients (the intelligent medical platform provides simple and clear health guidance and disease explanation to help patients understand their health status and make preliminary health decisions), and health maintenance and prevention demand groups; in the present embodiment, the intelligent medical platform runs based on the input of the target objects, that is, the target objects trigger a query operation (click the integrated modules) in the interface (for example, a webpage) of the query system (corresponding to the intelligent medical platform) and input query conditions or questions, interact with the intelligent medical platform based on the prompting project (that is, guide the user to fill in the relevant information through a friendly user interface), and thus obtain the query request information (corresponding to the input data) input by the target objects; it can be understood that in the process of the intelligent medical platform interacting with the target objects based on the prompting project, the intelligent medical platform can also know the query category label information (corresponding to determining one of case screening, medical popular science, information diagnosis, and consultation interaction) selected by the target object and the intention demand information required by the target object to be queried and obtained, for example, when the target object is a medical practitioner, and a symptom description of a target patient is input during the interaction with the intelligent medical platform, the intelligent medical platform is expected to provide the best treatment plan for the target patient (the intention demand information expected to be obtained by the medical practitioner)

[0024] Step S202, based on the query category label information, at least a natural language processing (NLP) algorithm is used to obtain candidate medical information associated with the intention demand information in a target database.

[0025] In the embodiment, after obtaining the query request information including the target object input, the intelligent medical platform connects with the target database (medical database, medical textbook, medical journal, case database, medical knowledge base, drug knowledge base) through the API interface, obtains relevant case data and medical knowledge, and gathers information from medical institutions, medical colleges and medical databases, that is, obtains candidate medical information associated with the intended demand information. In the embodiment, according to the query request information, the candidate medical information matching the intended demand information is preliminarily screened out in the target database by using the NLP technology.

[0026] Step S203, in the preset matching rule, determine the target matching rule corresponding to each query operation, and according to the target matching rule, screen out the target medical information corresponding to the intended demand information from the candidate medical information, and take the target medical information as the query result.

[0027] In the embodiment, after obtaining the candidate medical information, the target medical information is screened out from the candidate medical information again by using the preset matching rule, for example, when the case is screened, after obtaining the treatment scheme (corresponding to the candidate medical information) corresponding to the relevant case similar to the symptoms of the patient input by the target object, the optimal treatment scheme is selected from the screened treatment schemes according to the individual characteristics of the patient; for another example, when medical popularization is carried out, after the target object inputs the medical keywords, all medical items containing the medical keywords are preliminarily screened out, and based on the rule of selecting the core data of the drug, irrelevant literature or non-core items are excluded to obtain the target medical information meeting the requirement expectation.

[0028] Through the above steps S201 to S203, after responding to the query operation triggered by the target object in the query system, the query request information input by the target object is obtained based on the prompting engineering, the query request information includes the intended demand information and the query category label information; based on the query category label information, at least the natural language processing NLP algorithm is used to obtain the candidate medical information associated with the intended demand information in the target database; in the preset matching rule, determine the target matching rule corresponding to each query operation, and according to the target matching rule, screen out the target medical information corresponding to the intended demand information from the candidate medical information, and take the target medical information as the query result, solve the problem that the efficiency and accuracy of obtaining medical information are low in the scheme of providing medical information by using a large language model in the related technology, realize providing personalized medical knowledge and suggestions for different user groups, improve the efficiency and accuracy of information acquisition, in the environment with limited resources, medical information can be conveniently obtained, the accessibility and convenience of medical services are improved, and the diversified demand of users for medical information acquisition is met.

[0029] It should be noted that the smart medical platform of the embodiment of the present application provides unified medical information management and services for medical practitioners, medical students and patients by setting multiple modules including case screening, medical popularization, information diagnosis and consultation interaction; in the embodiment, through the prompting engineering technology, the smart medical platform can provide personalized medical knowledge and suggestions for different user groups, significantly improving the efficiency and accuracy of information acquisition; in a resource-limited environment, the target object can obtain medical information at any time and anywhere through the smart medical platform, improving the accessibility and convenience of medical services; in the embodiment, the smart medical platform of the embodiment of the present application provides diversified interactive modes for different user groups, medical practitioners can obtain detailed case analysis and diagnosis suggestions through the case screening module; medical students can learn basic medical knowledge through the medical popularization module, and carry out multi-round dialogue with the large language model through the interactive module to consolidate the learned content; patients can obtain simple and clear health advice through the preliminary diagnosis module; the smart medical platform makes the transmission of medical information more in-depth and rich through multi-level interaction and learning experience, meeting the diversified needs of users; in the embodiment, through the prompting engineering technology of the large language model, the smart medical platform can effectively integrate and utilize medical resources, provide timely and accurate medical advice and diagnosis, and for medical staff and patients with limited resources, the platform can help them make full use of scattered time for learning and health management, improving the efficiency of medical education and medical services.

[0030] In some embodiments, in response to the target object triggering the query operation of the query system, the query request information input by the target object is obtained based on the prompting engineering, and the following steps are implemented:

[0031] Step 21, after receiving the query operation triggered by the target object, the pre-trained large language model is used to carry out multi-round interaction with the corresponding target object through the prompting engineering, and the response information of the target object to the prompting engineering in each round of interaction is obtained.

[0032] Step 22, according to the response information, the intention demand information input by the target object is determined.

[0033] Step 23, the query category label corresponding to the query operation is determined, and the query request information is generated according to the response information of the target object, the intention demand information and the query category label.

[0034] In the embodiment, the corresponding prompt sentence is generated by using the prompt engineering of the large language model, and the corresponding prompt sentence is presented on the interface of the query system. Then, the target object fills in the corresponding response information on the interface according to the prompt sentence. After that, the intelligent medical platform determines the identity of the target object and the corresponding target information (for example: medical practitioners, medical students) according to the query operation triggered by the target object. In this way, the query category label corresponding to the query operation is determined, and the corresponding intention demand information is determined according to the response information, that is, the medical information intended to be queried by the target object is determined. Finally, the query request information is generated based on the query category label, the intention demand information and the response information. It can be understood that after the target object triggers the corresponding query operation, the query category label is generated, and the corresponding response information is classified. Therefore, when the intelligent medical platform receives the query request information, the target object currently performing the query request and the intended data to be obtained or queried can be quickly determined. Then, the intelligent medical platform can obtain the related medical information from the corresponding target database according to the response information in the query request information.

[0035] In the above steps, after receiving the query operation triggered by the target object, the pre-trained large language model is used for prompt engineering and multi-round interaction with the corresponding target object, and the response information of the target object to the prompt engineering in each round of interaction is obtained. According to the response information, the intention demand information input by the target object is determined, the query category label corresponding to the query operation is determined, and the query request information is generated according to the response information, the intention demand information and the query category label of the target object. The input data input by the target object is obtained, that is, the query request information is obtained.

[0036] In some embodiments, based on the query category label information, at least a natural language processing (NLP) algorithm is used to obtain candidate medical information associated with the query demand information in the target database, which is achieved by the following steps:

[0037] Step 31, after determining the corresponding query operation according to the query category label information, the corresponding response information is obtained from the corresponding query request information.

[0038] In the embodiment, the intelligent medical platform determines the identity of the target object and the intention demand information corresponding to the response information according to the query category label information representing the corresponding query operation.

[0039] Step 32, detecting candidate medical feature parameters associated with the first medical feature parameter in the target database using a natural language processing (NLP) algorithm, screening an intended medical feature parameter from the candidate medical feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameter, and taking the screened intended medical feature parameter as the candidate medical information, wherein the first medical feature parameter is generated by preprocessing the corresponding response information corresponding to the query operation, and the preprocessing at least includes one of the following: data cleaning, term standardization, key feature extraction, normalization, and image enhancement processing.

[0040] In this embodiment, after obtaining the response information input by the target object (as input data), the response information is preprocessed first, and then the corresponding first medical feature parameter is generated, for example: the symptom description input by the target object is data cleaned and term standardized, thereby generating the corresponding symptom code of the symptom description.

[0041] In this embodiment, after generating the first medical feature data, at least NLP technology is used to detect corresponding candidate medical feature parameters in the target database, for example: searching for similar cases in the case database according to the symptom description, and for example: comparing the symptoms in the case database with the keywords corresponding to the symptom code through NLP technology to find similar cases.

[0042] In this embodiment, after screening the candidate medical feature parameters corresponding to the first medical feature parameter, the corresponding feature analysis and screening are performed to obtain the intended medical feature parameter.

[0043] In some embodiments, detecting candidate medical feature parameters associated with the first medical feature parameter in the target database, screening an intended medical feature parameter from the candidate medical feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameter, is achieved by the following steps:

[0044] Step 41, in the case of determining that the query operation is case screening, using a BERT natural language processing model to perform semantic understanding processing on the first case data corresponding to the response information to obtain first disease data.

[0045] Step 42, using a BERT natural language processing model to detect candidate disease data associated with the first disease data in the medical database as the target database, and obtaining intended disease data from the candidate disease data according to the overlap degree between the candidate disease data and the first disease data, wherein the overlap degree is used to represent the disease similarity of the corresponding disease data.

[0046] Step 43, after determining the second case data associated with the intention disease data, the target historical medical information is screened from the historical medical information corresponding to the second case data by using the collaborative filtering algorithm based on neighborhood and matrix decomposition, to obtain the candidate medical information, wherein the historical medical information is used to represent the treatment scheme for the disease in the second case data, and the candidate medical information is used to represent the preliminary recommended scheme for the disease in the first case data.

[0047] In some embodiments, in the preset matching rule, the target matching rule corresponding to each query operation is determined, and the target medical information corresponding to the intention demand information is screened from the candidate medical information according to the target matching rule, including the following steps:

[0048] Step 51, determine the medical user portrait of the first object associated with the first case data, wherein the first object is used to represent the treatment object corresponding to the first case data, and the medical user portrait includes a plurality of dimensions of user characteristics, and the user characteristics include one of the following: age, gender, physiological parameters, and medical history parameters.

[0049] Step 52, according to the weighted weight corresponding to all user characteristics of the first object, the plurality of target historical medical information is screened from all target historical medical information corresponding to the candidate medical information.

[0050] Step 53, the trained treatment rate prediction model is used to process the plurality of target historical medical information screened, and according to the prediction treatment rate corresponding to each target historical medical information, the target historical medical information with the maximum prediction treatment rate is selected to obtain the target medical information, wherein the treatment rate prediction model is a prediction model trained based on one of random forest algorithm and extreme gradient decision tree.

[0051] In some optional embodiments, when the query operation is case screening, the target medical information is obtained by the following steps:

[0052] Step 1, the target object inputs the symptom description of the first object.

[0053] In this embodiment, the symptom description of the first object includes age, gender, chief complaint (such as persistent chest pain, dyspnea), medical history and current medication, etc. For example, a 30-year-old patient describes persistent chest pain.

[0054] Step 2, pre-process the symptom description.

[0055] In this embodiment, the symptom description is pre-processed, such as data cleaning and term standardization; specifically, "a 30-year-old patient describes persistent chest pain" is truncated as age: the patient's age (30 years old) is extracted; gender: identified according to the gender input by the patient (male); chief complaint: "persistent chest pain" is standardized as a medical term, such as "persistent chest pain" may be converted to "persistent chest pain" or "chest pain (persistent)"; remove redundant information: if there is redundant or irrelevant content (such as repeated description) in the patient's description; convert to standard terminology: map "chest pain" to standardized disease codes using a medical terminology set (such as ICD-10 or SNOMED-CT).

[0056] Step 3, using data visualization and NLP technology, search for similar cases in the case database according to the patient's symptoms, and return relevant results according to the overlap rate.

[0057] In this embodiment, by identifying the keyword "persistent chest pain" of the symptom, and comparing it with the symptoms in the case database through NLP technology, similar cases are found, for example, some cases where the patient also reports "chest pain" or "persistent chest pain", possible associated symptoms, whether accompanied by dyspnea, palpitations, sweating, etc.; then, according to the age, gender, duration of symptoms, etc. Information, filter out relevant cases, and return the best matching results according to the overlap degree (symptom similarity).

[0058] Step 4, according to the first object's medical history and current medication, screen out possible related treatment plans.

[0059] In this embodiment, according to the first object's medical history and current medication, the symptoms are analyzed and screened to find the most likely cause.

[0060] Step 5, use BERT natural language processing model for semantic understanding of symptoms, and vectorize the input symptoms.

[0061] In this embodiment, One-Hot encoding is used when vectorizing symptoms; in this embodiment, non-numerical data such as disease names, drug names, and symptoms are also encoded, and at the same time, each category of disease or drug can be converted to a binary feature through One-Hot encoding (such as "myocardial ischemia" can be [1, 0, 0], while "diabetes" is [0, 1, 0]).

[0062] Step 6, based on the semantic similarity of the vectorized encoding, perform symptom matching in the case database to filter out a more accurate set of cases.

[0063] Step 7, using the treatment history of patients with similar symptoms as a preliminary recommendation, screening the treatment plan used by patients with similar symptoms through a collaborative filtering algorithm (such as a collaborative filtering algorithm based on neighborhood and matrix decomposition).

[0064] In this embodiment, according to the individual characteristics of the first object such as age, gender and constitution, the weight corresponding to each individual characteristic is adjusted when screening the treatment plan; for example, older patients may not be suitable for certain drugs or doses, so the weight of the age factor in the model is adjusted to make the treatment plan more suitable for the needs of a specific population, and for example: through the drug knowledge base (such as DrugBank) to detect the interaction risk between the recommended drugs and the current medication, remove the treatment plan with potential conflicts; also through a classification model such as random forest or XGBoost, according to the patient's characteristics and the treatment results of patients with similar symptoms to predict the success rate of the treatment plan, and screen out the treatment plan with the best effect, and also

[0065] Step 8, generating a personalized treatment recommendation for the screened treatment plan.

[0066] In some embodiments, detecting alternative medical feature parameters associated with the first medical feature parameter in the target database, screening the intended medical feature parameter from the alternative feature parameters according to the similarity between the alternative medical feature parameters and the first medical feature parameter, and achieving the following steps:

[0067] Step 61, in the case of determining that the query operation is medical popular science, searching for the target data entry corresponding to the medical keyword in the corresponding preset word table of the medical database as the target database, wherein the medical database stores a plurality of medical popular science text data, and the medical popular science text data includes at least one medical data entry.

[0068] In this embodiment, after the target object inputs the medical keyword as the response information, the intelligent medical platform will determine the target data entry associated with the medical keyword based on the input medical keyword according to the corresponding word table, for example: when the medical keyword is a drug name, the corresponding associated target data entry includes: alias, generic name, belonging, purpose, source, manufacturer; In this embodiment, the source of the target medical information obtained is the medical popular science text data stored in the medical database, and each medical popular science text data is composed of a medical data entry and a corresponding vocabulary.

[0069] Step 62, using the preset target NLP method, detecting the target data item in each medical popular science text data corresponding to the medical data item, and retrieving the medical popular science text data including at least one target data item as the candidate medical popular science text data, obtaining the intended medical feature parameter including the candidate medical popular science text data, wherein the target NLP method includes one of the following: forward maximum matching method, reverse maximum matching method, bidirectional scanning method and word-by-word traversal method, and the similarity is determined according to the ratio of the number of target data items and medical data items in the medical popular science text data.

[0070] In this embodiment, in all medical popular science text data stored in the medical database, the target NLP method is used for retrieval, and medical popular science text data containing all target data items are screened out, and the screened medical popular science text data are used as the intended medical feature parameter.

[0071] In some embodiments, in the preset matching rule, the target matching rule corresponding to each query operation is determined, and the target medical information corresponding to the intended demand information is screened out from the candidate medical information according to the target matching rule, including the following steps:

[0072] Step 71, detecting non-core items in all medical data items corresponding to the candidate medical popular science text data.

[0073] Step 72, after deleting the candidate medical popular science text data in which the non-core items are detected, the remaining candidate medical popular science text data is used as the target medical popular science text data corresponding to the medical keywords, and the target medical popular science text data is used as the target medical information corresponding to the intended demand information.

[0074] In some optional embodiments, when the query operation is medical popular science, the target medical information is obtained by the following steps:

[0075] Step 1, the target object inputs the keywords (drug name or its alias, generic name, etc.).

[0076] Step 2, after the intelligent medical platform receives the keywords, it performs preliminary retrieval in the medical database based on the keywords using NLP technology, and screens out medical popular science text data containing all target data items of the drug.

[0077] In this embodiment, the target data item includes one of the following: drug ingredients, product instructions, clinical research, and drug interaction data.

[0078] In this embodiment, the smart medical platform will first extract the trade name, generic name and chemical name of the drug from the target data entries related to the drug, confirm whether the drug has common aliases, and further match them. In this embodiment, the search includes but is not limited to searching for the therapeutic efficacy and mechanism of action of the drug, analyzing its effect on specific diseases or symptoms, and according to its main ingredients and mechanism, sorting out its main indications and secondary indications; search for drug interaction information, find drugs or foods that have incompatibility with it; summarize the possible synergistic drugs of the drug, list which drugs are compatible to enhance efficacy; search for research literature of the drug, judge its actual application value in clinical practice; combine the mechanism of action of the drug, sort out its unique value in medicine and health management; in this embodiment, according to the drug instructions and research data, the drug's applicable population (such as adults, children, the elderly, pregnant women, etc.) is determined, and through side effect and contraindication information, the population that is not suitable for use is screened out; extract detailed information such as recommended dose, medication frequency, and medication time from product instructions or medical recommendations; compare whether different methods of taking the medicine have a significant impact on drug efficacy, and record adjustment suggestions in special cases. Find the known side effects of the drug and summarize common and rare adverse reactions.

[0079] In some optional embodiments, the target object inputs a keyword, for example, the target object inputs "aspirin", and the smart medical platform receives the drug name and searches and matches in the medical database to screen out entries such as "alias: acetylsalicylic acid", "belongs to: penicillin antibiotics", "purpose: treat various infections caused by sensitive bacteria, etc.", "source: it is originally derived from willow bark and began to be synthesized artificially in the late 19th century", "manufacturer: may include multiple pharmaceutical companies.

[0080] Step 3, further select the core data of the drug from the screened target data entries, and exclude irrelevant literature or non-core entries.

[0081] In this embodiment, the smart medical platform further screens all relevant target data entries that have been screened out, and excludes non-core entries such as irrelevant literature such as "manufacturer" and "shelf life".

[0082] In this embodiment, the core information includes composition description, mechanism of action; in this embodiment, among the screened data, each type of information is processed specifically to gradually obtain the required target information, specifically, the trade name, generic name and chemical name of the drug are extracted from the drug-related entries; it is confirmed whether the drug has common aliases; the therapeutic efficacy and mechanism of action of the drug are searched, and its effect on specific diseases or symptoms is analyzed; according to its main ingredients and mechanism, its main indications and secondary indications are sorted out; the interaction information of the drug is searched to find drugs or foods that are incompatible with it; the possible synergistic drugs of the drug are summarized, and it is listed with which drugs are generated to enhance efficacy; the research literature of the drug is searched to judge its actual application value in the clinic; combined with the mechanism of action of the drug, its unique value in medicine and health management is sorted out; according to the drug instruction and research data, the suitable population of the drug (such as adults, children, the elderly, pregnant women, etc.) is clarified; through the side effect and contraindication information, the population who is not suitable for use is screened out; the recommended dose, medication frequency, medication time and other detailed information are extracted from the product instruction or medical advice; whether different taking methods have a significant effect on drug efficacy is compared, and adjustment suggestions under special circumstances are recorded; the known side effects of the drug are searched, and common and rare adverse reactions are summarized; the side effects are classified according to the severity, and the precautions for avoiding adverse reactions are attached.

[0083] Step 4, the smart medical platform organizes and displays these information in text, pictures, videos, etc. for medical personnel to refer or provide clear medication guidance for patients.

[0084] In this embodiment, when displaying drug information, the smart medical platform will also recommend relevant medical articles, research papers or expert lectures according to user portraits, helping users to understand drugs and related knowledge more deeply; in this embodiment, the smart medical platform has an interactive question and answer area, where users can ask questions about drugs or other medical issues, which will be answered by medical experts or professional certified volunteers on the smart medical platform; in this embodiment, the smart medical platform obtains a large amount of medical basic knowledge from medical textbooks, medical journals, medical databases, etc.; the obtained data is structured and processed using NLP technology to ensure accurate matching of related information when users query; users input the queried medical knowledge or questions on the smart medical platform, and the smart medical platform will search and match in the medical knowledge base according to the query conditions.

[0085] In some embodiments, candidate medical feature parameters associated with the first medical feature parameter are detected in the target database, and the intended medical feature parameter is selected from the candidate medical feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameter, which is achieved by the following steps:

[0086] Step 81, in the case where the query operation is information diagnosis and the response information includes medical detection data, the pre-constructed convolutional neural network (CNN) is used to process the pre-processed medical detection data to generate first medical feature parameters, wherein the first medical feature parameters include diagnosis text data and / or detection images, and the pre-processing includes data cleaning, normalization, key feature extraction, and image enhancement processing.

[0087] Step 82, using a knowledge graph-based natural language processing (NLP) algorithm, the candidate medical feature parameters corresponding to the historical detection data are matched and analyzed with the first medical feature parameters to determine the similarity of the candidate medical feature parameters and the corresponding first medical feature parameters, wherein the historical detection data includes existing medical detection data included in a medical database as a target database, and the candidate medical feature parameters include first diagnosis text data and / or first detection images corresponding to the historical detection data.

[0088] Step 83, selecting the candidate medical feature parameters with a similarity greater than a preset threshold, and taking the selected candidate medical feature parameters as intended medical feature parameters.

[0089] In some embodiments, in the preset matching rule, the target matching rule corresponding to each query operation is determined, and according to the target matching rule, the target medical information corresponding to the intended demand information is selected from the candidate medical information, including the following steps: obtaining the candidate medical feature parameters as the intended medical feature parameters, and updating the first medical feature parameters with the corresponding candidate medical feature parameters to generate target medical information including target medical feature parameters, wherein the target medical feature parameters include target diagnosis text data and / or target detection images.

[0090] In some optional embodiments, when the query operation is information diagnosis, the target medical information is obtained by the following steps:

[0091] Step 1, the target object inputs the preliminary diagnosis result and related examination data of a patient.

[0092] In this embodiment, the preliminary diagnosis result and the related examination data correspond to the response information.

[0093] Step 2, after receiving the medical detection data corresponding to the response information, the intelligent medical platform verifies and checks to ensure the accuracy and integrity of the medical detection data.

[0094] Step 3, the intelligent medical platform uses a convolutional neural network (CNN) to analyze the preliminary diagnosis results and related examination data, returns the verification results of the preliminary diagnosis results, and further diagnosis suggestions based on the test images (corresponding to electrocardiogram and blood test results). If the intelligent medical platform finds that the preliminary diagnosis has errors or omissions, it will provide correction suggestions or additional examination suggestions.

[0095] In this embodiment, the intelligent medical platform first performs detailed data preprocessing on the data input by the target object, including data cleaning, normalization, and then accurately extracts key features. The extracted features are used to match and analyze the case data widely collected in the medical database. When the similarity is higher than the preset threshold, it means that the current case has significant similarity with some cases in the database. These similar cases can be used as valuable references.

[0096] In this embodiment, the target object can also use the scanning function provided by the intelligent medical platform to detect medical diagnosis images. The intelligent medical platform performs edge detection, noise removal, and image enhancement processing on the medical diagnosis images, and compares the processed medical diagnosis images with the normal images stored in the medical database. Any significant difference area found in the comparison will be marked in a prominent form (such as highlighting, labeling, etc.), and the analysis results will be fed back together.

[0097] In some embodiments, candidate medical feature parameters associated with the first medical feature parameter are detected in the target database, and the intended medical feature parameter is selected from the candidate feature parameters according to the similarity between the candidate medical feature parameter and the first medical feature parameter, which is achieved by the following steps:

[0098] Step 91, in the case of determining that the query operation is a consultation interaction, using a BERT natural language processing model to perform semantic understanding processing on the preprocessed inquiry text data to obtain semantic features corresponding to the inquiry text, wherein the preprocessing includes data cleaning, and the response information includes inquiry text data.

[0099] Step 92, in the preset medical database, using semantic features and natural language generation (NLG) methods to generate a plurality of response text data, wherein the candidate medical feature parameters include response text data, and the response text data is used to represent the response to the consultation question of the target object associated with the inquiry text data in the consultation interaction process.

[0100] Step 93, according to the actual demand information determined by analyzing the semantic features, calculating the matching degree of each response text data and the actual demand information, and taking the matching degree as the corresponding similarity, and selecting at least one response text data as the intended medical feature parameter from the plurality of response text data.

[0101] In some embodiments, in the preset matching rule, the target matching rule corresponding to each query operation is determined, and the target medical information corresponding to the intention demand information is screened from the candidate medical information according to the target matching rule, including the following steps: selecting the response text data with the maximum matching degree from the screened at least one response text data as the target medical information corresponding to the intention demand information.

[0102] In this embodiment, the wisdom medical platform retrieves and matches in the medical knowledge base according to the user's question and health information through natural language processing, prompting engineering technology, large language model, etc., and provides the user with corresponding health advice and preliminary diagnosis; the user inputs the consultation question through the chat window, and the wisdom medical platform outputs the corresponding health advice and preliminary diagnosis result in real time, so as to adjust the user's recent physical condition.

[0103] In some optional embodiments, when the query operation is a consultation interaction, the target medical information is obtained by the following steps:

[0104] Step 1, the target object inputs a health problem through the wisdom medical platform.

[0105] Step 2, after the wisdom medical platform receives the natural language input of the user, semantic understanding and analysis are performed, and natural language generation technology and a medical database are used to generate answers and suggestions for the user's question.

[0106] In this embodiment, the wisdom medical platform receives the text data corresponding to the natural language input of the user, and uses natural language processing (NLP) technology to deeply analyze the user's question through a large language data model to extract effective information.

[0107] For example, when the user asks “I feel very tired recently, how should I improve my diet and work and rest?”, the effective information is accurately identified as “tiredness”, “diet” and “work and rest”. After receiving this effective information, the platform immediately screens and combines in the pre-constructed database to find relevant solutions.

[0108] In this embodiment, the NLP technology processing flow specifically includes: first, the natural language text data is carefully screened and preprocessed to remove irrelevant information and retain key content; second, the NLP model is used to deeply analyze the preprocessed text and convert it into a structured data form, which can capture the semantic features and emotional tendencies in the text; third, the obtained structured data is parsed, and the actual demand information of the user, i.e. the intention demand information, is identified and determined through algorithm logic, including but not limited to the user's specific symptoms, areas of concern and expected solutions.

[0109] Step 3: The intelligent medical platform provides relevant medical knowledge and health tips to help users better understand and manage their health status.

[0110] In some embodiments, after the target medical information corresponding to the intended demand information is screened out, the method further comprises: generating target recommendation information based on the target medical information, and displaying the target recommendation information in a preset display format on the interface of the query system, wherein the display format includes one of the following: text display, picture display, and video image display.

[0111] The embodiment also provides a medical information acquisition device based on the prompting engineering, which is used to implement the above-mentioned embodiments and preferred embodiments, and details are not repeated. As used below, the terms "module", "unit", "sub-unit", and the like can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and contemplated.

[0112] Figure 3 is a structural block diagram of medical information acquisition based on the prompting engineering according to the embodiment of the application, as shown in Figure 3 The device includes an acquisition module 31, a screening module 32, and a processing module 33, wherein

[0113] The acquisition module 31 is configured to acquire the query request information input by the target object based on the prompting engineering in response to the query operation triggered by the target object on the query system, wherein the query request information includes the intended demand information and the query category label information, and the query category label information is used to represent one of the following query operations: case screening, medical popularization, information diagnosis, and consultation interaction.

[0114] The screening module 32 is coupled to the acquisition module 31 and is configured to acquire candidate medical information associated with the intended demand information in the target database based on the query category label information and at least using a natural language processing (NLP) algorithm.

[0115] The processing module 33 is coupled to the screening module 32 and is configured to determine a target matching rule corresponding to each query operation in a preset matching rule, and screen out target medical information corresponding to the intended demand information from the candidate medical information according to the target matching rule, and take the target medical information as a query result.

[0116] In some embodiments, the acquisition module 31 further includes:

[0117] An interaction unit is configured to perform multi-round interaction with the target object by using the prompt engineering of the pre-trained large language model and obtain response information of the target object to the prompt engineering after receiving a query operation triggered by the target object.

[0118] A determination unit is coupled to the interaction unit and configured to determine the intention demand information input by the target object according to the response information.

[0119] A generation unit is coupled to the determination unit and configured to determine a query category label corresponding to the query operation, and generate query request information according to the response information, the intention demand information and the query category label of the target object.

[0120] In some embodiments, the screening module 32 further includes:

[0121] An acquisition unit is configured to acquire the corresponding response information from the corresponding query request information after determining the corresponding query operation according to the query category label information.

[0122] A selection unit is coupled to the acquisition unit and configured to detect candidate medical feature parameters associated with the first medical feature parameter in the target database by using a natural language processing (NLP) algorithm, screen the intended medical feature parameter from the candidate medical feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameter, and select the screened intended medical feature parameter as the candidate medical information, wherein the first medical feature parameter is generated by preprocessing the response information corresponding to the query operation, and the preprocessing at least includes one of the following: data cleaning, term standardization, key feature extraction, normalization and image enhancement processing.

[0123] In some embodiments, the selection unit is further configured to, in a case where it is determined that the query operation is case screening, perform semantic understanding processing on first case data corresponding to the response information by using a BERT natural language processing model to obtain first disease data; detect candidate disease data associated with the first disease data in a medical database serving as the target database by using the BERT natural language processing model, and acquire intended disease data from the candidate disease data according to the overlap degree between the candidate disease data and the first disease data, wherein the overlap degree is used to represent the disease similarity of the corresponding disease data; after determining second case data associated with the intended disease data, screen target historical medical information from historical medical information corresponding to the second case data by using a collaborative filtering algorithm based on neighborhood and matrix decomposition to obtain the candidate medical information, wherein the historical medical information is used to represent a scheme for treating a disease in the second case data, and the candidate medical information is used to represent a preliminary recommended scheme for treating a disease in the first case data.

[0124] In some embodiments, the processing module 33 is further configured to determine a medical user portrait of a first object associated with the first case data, wherein the first object is used to represent a treatment object corresponding to the first case data, the medical user portrait comprises a plurality of dimensions of user features, and the user features comprise one of the following: age, gender, physiological parameters, and medical history parameters; according to a weighted weight corresponding to all user features of the first object, a plurality of target historical medical information from all target historical medical information corresponding to the candidate medical information is screened; the trained treatment rate prediction model is used to process the plurality of target historical medical information, and according to the predicted treatment rate corresponding to each target historical medical information, the target historical medical information with the maximum predicted treatment rate is selected to obtain the target medical information, wherein the treatment rate prediction model is a prediction model trained based on one of a random forest algorithm and an extreme gradient decision tree.

[0125] In some embodiments, the selecting unit is further configured to, in a case where it is determined that the query operation is a medical popular science, search for a target data entry corresponding to a medical keyword of the response information in a preset vocabulary corresponding to a medical database serving as a target database, wherein the medical database stores a plurality of medical popular science text data, and each medical popular science text data comprises at least one medical data entry; detect the target data entry in each medical data entry of the medical popular science text data by using a preset target NLP method, and select medical popular science text data comprising at least one target data entry as candidate medical popular science text data to obtain an intended medical feature parameter comprising the candidate medical popular science text data, wherein the target NLP method comprises one of the following: a forward maximum matching method, a reverse maximum matching method, a bidirectional scanning method, and a word-by-word traversal method, and the similarity is determined according to a ratio of the number of target data entries to the number of medical data entries in the medical popular science text data.

[0126] In some embodiments, the processing module 33 is further configured to detect a non-core entry in all medical data entries corresponding to the candidate medical popular science text data; after deleting the candidate medical popular science text data in which the non-core entry is detected, the remaining candidate medical popular science text data is used as target medical popular science text data corresponding to the medical keyword, and the target medical popular science text data is used as the target medical information corresponding to the intended demand information.

[0127] In some embodiments, the selecting unit is further configured to, in a case where the query operation is determined to be information diagnosis and the response information comprises medical detection data, process the pre-processed medical detection data using a pre-constructed convolutional neural network (CNN) to generate first medical feature parameters, wherein the first medical feature parameters comprise diagnosis text data and / or detection images, and the pre-processing comprises data cleaning, normalization, key feature extraction, and image enhancement processing; perform matching analysis on the candidate medical feature parameters corresponding to historical detection data and the first medical feature parameters using a knowledge graph-based natural language processing (NLP) algorithm to determine the similarity between the candidate medical feature parameters and the corresponding first medical feature parameters, wherein the historical detection data comprises existing medical detection data included in a medical database serving as the target database, and the candidate medical feature parameters comprise first diagnosis text data and / or first detection images corresponding to the historical detection data; and select the candidate medical feature parameters with a similarity greater than a preset threshold as the intended medical feature parameters.

[0128] In some embodiments, the processing module 33 is further configured to obtain the candidate medical feature parameters as the intended medical feature parameters, and update the first medical feature parameters with the corresponding candidate medical feature parameters to generate target medical information comprising target medical feature parameters, wherein the target medical feature parameters comprise target diagnosis text data and / or target detection images.

[0129] In some embodiments, the selecting unit is further configured to, in a case where the query operation is determined to be consultation interaction, perform semantic understanding processing on the pre-processed inquiry text data using a BERT natural language processing model to obtain semantic features corresponding to the inquiry text, wherein the pre-processing comprises data cleaning, and the response information comprises inquiry text data; in a preset medical database, generate a plurality of response text data using the semantic features and a natural language generation (NLG) method, wherein the candidate medical feature parameters comprise the response text data, and the response text data are used to represent responses to questions consulted by a target object associated with the inquiry text data in a consultation interaction process; calculate the matching degree of each response text data and the actual demand information determined by analyzing the semantic features, and use the matching degree as the corresponding similarity, and select at least one response text data from the plurality of response text data as the intended medical feature parameters.

[0130] In some embodiments, the processing module 33 is further configured to select the response text data with the largest matching degree from the at least one selected response text data as the target medical information corresponding to the intended demand information.

[0131] In some embodiments, the medical information query device is further configured to, after screening out the target medical information corresponding to the intention demand information, generate target recommendation information based on the target medical information, and display the target recommendation information in a preset display format on an interface of the query system, wherein the display format comprises one of the following: text display, picture display, and video image display.

[0132] The embodiment also provides an electronic device including a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to perform the steps in any of the above method embodiments.

[0133] Optionally, the electronic device can further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0134] Optionally, in the embodiment, the processor can be configured to execute the following steps through the computer program:

[0135] S1, in response to a query operation triggered by a target object on a query system, obtaining query request information input by the target object based on a prompt engineering, wherein the query request information includes intention demand information and query category label information, and the query category label information is used to represent one of the following query operations: case screening, medical popularization, information diagnosis, and consultation interaction.

[0136] S2, based on the query category label information, at least using a natural language processing (NLP) algorithm to obtain candidate medical information associated with the intention demand information in a target database.

[0137] S3, in a preset matching rule, determining a target matching rule corresponding to each query operation, and according to the target matching rule, screening out target medical information corresponding to the intention demand information from the candidate medical information, and taking the target medical information as a query result.

[0138] In addition, in combination with the medical information query method based on the prompt engineering in the above embodiments, the embodiment of the application can provide a storage medium for implementation. The storage medium stores a computer program; when the computer program is executed by a processor, any one of the medical information query methods based on the prompt engineering in the above embodiments is implemented.

[0139] Those skilled in the art should understand that each technical feature of the above embodiments can be combined arbitrarily, and in order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist, they should be considered as the scope of the description.

[0140] The above embodiments only express several implementation ways of the present application, and the description is specific and detailed, but it should not be understood as a limitation to the patent scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A medical information search method based on a hint engineering, characterized by, Comprise: In response to the target object in the query system triggered query operation, based on the prompt engineering to obtain the target object input query request information, wherein the query request information includes the intention demand information and the query category label information, the query category label information is used to represent one of the following query operations: case screening, medical popular science, information diagnosis and consultation interaction; Based on the query category label information, at least using natural language processing NLP algorithm, in the target database, the candidate medical information associated with the intention demand information is obtained; In the preset matching rule, determine the target matching rule corresponding to each query operation, and according to the target matching rule, from the candidate medical information, the target medical information corresponding to the intention demand information is screened out, and the target medical information is taken as the query result; Wherein, based on the query category label information, at least using natural language processing NLP algorithm, in the target database, the candidate medical information associated with the query demand information is obtained, including: After determining the corresponding query operation according to the query category label information, the corresponding response information is obtained from the corresponding query request information; Using the natural language processing NLP algorithm, the candidate medical feature parameters associated with the first medical feature parameters are detected in the target database, the intention medical feature parameters are screened from the candidate medical feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameters, and the screened intention medical feature parameters are taken as the candidate medical information, wherein the first medical feature parameters are generated by preprocessing the response information corresponding to the query operation, and the preprocessing at least includes one of the following: data cleaning, term standardization, key feature extraction, normalization, image enhancement processing; Wherein, in the target database, the candidate medical feature parameters associated with the first medical feature parameters are detected, the intention medical feature parameters are screened from the candidate feature parameters according to the similarity between the candidate medical feature parameters and the first medical feature parameters, including: In the case of determining that the query operation is medical popular science, in the corresponding preset word table of the medical database as the target database, the target data entry corresponding to the medical keyword corresponding to the response information is searched, wherein the medical database stores a plurality of medical popular science text data, and the medical popular science text data includes at least one medical data entry; detect the target data item in each of the medical popular science text data corresponding to the medical data item by using a preset target NLP method, and retrieve the medical popular science text data including at least one target data item as alternative medical popular science text data, to obtain the intended medical feature parameter including the alternative medical popular science text data, wherein the target NLP method includes one of the following: forward maximum matching method, reverse maximum matching method, bidirectional scanning method and word-by-word traversal method, and the similarity is determined according to the ratio of the number of target data items and medical data items in the medical popular science text data; and / or determine a target matching rule corresponding to each of the query operations, and according to the target matching rule, filter the target medical information corresponding to the intended demand information from the candidate medical information, including: detecting non-core items in all medical data items corresponding to the alternative medical popular science text data; after deleting the alternative medical popular science text data in which the non-core items are detected, taking the remaining alternative medical popular science text data as the target medical popular science text data corresponding to the medical keywords, and taking the target medical popular science text data as the target medical information corresponding to the intended demand information.

2. The method of claim 1, wherein, In response to the query operation triggered by the target object in the query system, the query request information input by the target object is obtained based on the prompt engineering, including: After receiving the query operation triggered by the target object, the pre-trained large language model is used to interact with the corresponding target object in multiple rounds based on the prompt engineering, and the response information of the target object to the prompt engineering in each round of interaction is obtained; According to the response information, the intended demand information input by the target object is determined; determine the query category label corresponding to the query operation, and generate the query request information according to the response information, the intended demand information and the query category label corresponding to the target object.

3. The method of claim 1, wherein, detect the alternative medical feature parameter associated with the first medical feature parameter in the target database, and filter the intended medical feature parameter from the alternative feature parameter according to the similarity between the alternative medical feature parameter and the first medical feature parameter, including: In the case of determining that the query operation is case screening, the first case data corresponding to the response information is processed by the BERT natural language processing model to obtain first disease data; using the BERT natural language processing model, detecting alternative disease data associated with the first disease data in the medical database as the target database, and according to the overlap between the alternative disease data and the first disease data, obtaining the intended disease data from the alternative disease data, wherein the overlap is used to represent the disease similarity of the corresponding disease data; After determining the second case data associated with the intention disease data, a target historical medical information is screened from historical medical information corresponding to the second case data by using a collaborative filtering algorithm based on neighborhood and matrix decomposition, to obtain the candidate medical information, wherein the historical medical information is used to represent a treatment scheme for a disease in the second case data, and the candidate medical information is used to represent a preliminary recommended scheme for treating a disease in the first case data.

4. The method of claim 3, wherein, In the preset matching rule, a target matching rule corresponding to each of the query operations is determined, and according to the target matching rule, target medical information corresponding to the intention demand information is screened from the candidate medical information, including: Determine the medical user portrait of the first object associated with the first case data, wherein the first object represents the treatment object corresponding to the first case data, and the medical user portrait includes user features in multiple dimensions, and the user features include one of the following: age, gender, physiological parameters, and medical history parameters; According to the weighted weight corresponding to all the user features of the first object, a plurality of target historical medical information is screened from all the target historical medical information corresponding to the candidate medical information; The trained treatment rate prediction model is used to process the plurality of target historical medical information screened, and according to the predicted treatment rate corresponding to each of the target historical medical information, the target historical medical information with the maximum predicted treatment rate is selected to obtain the target medical information, wherein the treatment rate prediction model is a prediction model trained based on one of a random forest algorithm and an extreme gradient decision tree.

5. The method of claim 1, wherein, In the target database, a candidate medical feature parameter associated with a first medical feature parameter is detected, and an intention medical feature parameter is screened from the candidate feature parameter according to the similarity between the candidate medical feature parameter and the first medical feature parameter, including: In the case where the query operation is information diagnosis and the response information includes medical detection data, the pre-processed medical detection data is processed by using a pre-constructed convolutional neural network (CNN) to generate the first medical feature parameter, wherein the first medical feature parameter includes diagnosis text data and / or detection image, and the preprocessing includes data cleaning, normalization, key feature extraction and image enhancement processing; The similarity between the candidate medical feature parameter and the first medical feature parameter corresponding to the historical detection data is determined by using a knowledge graph-based natural language processing (NLP) algorithm to perform matching analysis on the candidate medical feature parameter and the first medical feature parameter, wherein the historical detection data includes existing medical detection data included in a medical database as the target database, and the candidate medical feature parameter includes first diagnosis text data and / or first detection image corresponding to the historical detection data; The candidate medical feature parameter with a similarity greater than a preset threshold is selected as the intention medical feature parameter; and / or In the preset matching rule, the target matching rule corresponding to each query operation is determined, and the target medical information corresponding to the intended demand information is screened from the candidate medical information according to the target matching rule, including: obtaining the candidate medical feature parameter as the intended medical feature parameter, and updating the first medical feature parameter with the corresponding candidate medical feature parameter to generate the target medical information including the target medical feature parameter, wherein the target medical feature parameter includes target diagnosis text data and / or target detection image.

6. The method of claim 1, wherein, In the target database, the candidate medical feature parameter associated with the first medical feature parameter is detected, and the intended medical feature parameter is screened from the candidate feature parameter according to the similarity between the candidate medical feature parameter and the first medical feature parameter, including: In the case of determining that the query operation is a consultation interaction, the BERT natural language processing model is used for semantic understanding processing of the preprocessed inquiry text data to obtain the semantic features corresponding to the inquiry text, wherein the preprocessing includes data cleaning, and the response information includes the inquiry text data; In the preset medical database, the semantic features and natural language generation (NLG) method are used to generate a plurality of response text data, wherein the candidate medical feature parameter includes the response text data, and the response text data is used to represent the response to the consulted question of the target object associated with the inquiry text data in the consultation interaction process; According to the actual demand information determined by analyzing the semantic features, the matching degree of each response text data and the actual demand information is calculated, and the matching degree is used as the corresponding similarity to screen at least one response text data as the intended medical feature parameter from the plurality of response text data; and / or In the preset matching rule, the target matching rule corresponding to each query operation is determined, and the target medical information corresponding to the intended demand information is screened from the candidate medical information according to the target matching rule, including: selecting the response text data with the largest matching degree from the at least one screened response text data as the target medical information corresponding to the intended demand information.

7. The method according to any one of claims 3 to 6, characterized in that, After screening the target medical information corresponding to the intended demand information, the method further includes: generating target recommendation information based on the target medical information, and displaying the target recommendation information in a preset display format on the interface of the query system, wherein the display format includes one of the following: text display, picture display, video image display.

8. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the steps of the medical information query method based on the prompt engineering in any one of claims 1 to 7. The computer program is executed by the processor to realize the steps of the medical information query method based on the prompt engineering in any one of claims 1 to 7.

Images