Drug recommendation method and device, electronic equipment, storage medium and product

Abstract

The invention provides a drug recommendation method and device, electronic equipment, a storage medium and a product, and the method comprises the steps: predicting candidate recommended drugs based on user inquiry information input by a user and / or key information corresponding to the user inquiry information, and obtaining the drug information of the candidate recommended drugs from a drug knowledge base; matching medical knowledge corresponding to the user inquiry information from a pre-constructed medical knowledge base by utilizing the user inquiry information and the key information; and performing integrated reasoning on the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user inquiry information, and determining a drug recommendation result corresponding to the drug use inquiry information. According to the scheme, medicine recommendation can be automatically carried out, the medicine recommendation efficiency and accuracy are improved, after the candidate recommended medicine is predicted, the candidate recommended medicine can be further verified in combination with the medicine information and medical knowledge of the candidate recommended medicine, the final medicine recommendation result is determined, and the medicine recommendation efficiency is improved. And the accuracy of drug recommendation is further improved.

Description

Technical Field

[0001] The present application relates to the field of natural language processing technology, and in particular to a drug recommendation method, device, electronic device, storage medium and product. Background Art

[0002] In the field of healthcare, drug recommendation is an important link. Existing drug recommendation methods usually rely on doctors' professional knowledge and experience, and doctors recommend drugs to patients. However, in most cases, face-to-face communication with doctors is required, which makes drug recommendation inefficient and doctors have a large workload. Long working hours may affect the accuracy of drug recommendations. Summary of the invention

[0003] Based on the above needs, the present application proposes a drug recommendation method, device, electronic device, storage medium and product, which can reduce labor costs and improve the efficiency and accuracy of drug recommendations.

[0004] To achieve the above objectives, this application proposes the following technical solutions:

[0005] According to a first aspect of an embodiment of the present application, a drug recommendation method is provided, comprising:

[0006] Based on the user query information input by the user and / or the key information corresponding to the user query information, predict the candidate recommended drugs corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information;

[0007] Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base;

[0008] The drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the user query information.

[0009] Optionally, the medical knowledge base includes a medical expert opinion knowledge base, which stores medication opinions provided by medical experts for various diseases;

[0010] Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base includes:

[0011] The key information is matched with the medication opinions in the medical expert opinion knowledge base, and the medication opinions matched with the key information are used as the medical knowledge corresponding to the user query information.

[0012] Optionally, the medical knowledge base includes a medical field knowledge base, which stores medical field knowledge generated based on medical literature, clinical data, and network medical data;

[0013] Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base includes:

[0014] The user query information is matched with the medical field knowledge in the medical field knowledge base, and the medical field knowledge matched with the user query information is used as the medical knowledge corresponding to the user query information.

[0015] Optionally, integrating and reasoning the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information includes:

[0016] Based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information, the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the user query information.

[0017] Optionally, based on the respective priorities of the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information, integrating and reasoning the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information includes:

[0018] Constructing a knowledge integration reasoning instruction based on the priorities of the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information;

[0019] The user query information, the drug information of the candidate recommended drugs, the medical knowledge corresponding to the user query information and the knowledge integration reasoning instructions are input into a pre-trained first large language model to obtain a drug recommendation result corresponding to the user query information.

[0020] Optionally, based on the user query information input by the user and / or key information corresponding to the user query information, predicting the candidate recommended drugs corresponding to the user query information includes:

[0021] The user query information input by the user and / or the key information corresponding to the user query information, as well as the pre-constructed drug recommendation instructions, are input into the pre-trained second largest language model to obtain candidate recommended drugs corresponding to the user query information.

[0022] Optionally, after inputting the user query information input by the user and / or the key information corresponding to the user query information, and the pre-built drug recommendation instruction into the pre-trained second language model to obtain the candidate recommended drugs corresponding to the user query information, the method further includes:

[0023] The user query information, the drug information of the candidate recommended drugs and the pre-constructed matching instructions are input into the second largest language model, so that the second largest language model determines whether the candidate recommended drugs are consistent with the user query information according to the matching instructions, and deletes the candidate recommended drugs that do not match the user query information from all candidate recommended drugs.

[0024] According to a second aspect of an embodiment of the present application, a drug recommendation device is provided, comprising:

[0025] A drug prediction module, used to predict the candidate recommended drugs corresponding to the user query information based on the user query information input by the user and / or the key information corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information;

[0026] A matching module, used to match the medical knowledge corresponding to the user query information from a pre-built medical knowledge base using the user query information and the key information;

[0027] The integrated reasoning module is used to perform integrated reasoning on the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information.

[0028] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including: a memory and a processor;

[0029] The memory is connected to the processor and is used to store programs;

[0030] The processor is used to implement the above-mentioned drug recommendation method by running the program in the memory.

[0031] According to a fourth aspect of an embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned drug recommendation method is implemented.

[0032] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, comprising computer program instructions, which, when executed by a processor, enable the processor to implement the above-mentioned drug recommendation method.

[0033] The drug recommendation method proposed in the present application predicts the candidate recommended drugs corresponding to the user query information based on the user query information input by the user and / or the key information corresponding to the user query information, and obtains the drug information of the candidate recommended drugs from a pre-constructed drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information; using the user query information and the key information, the medical knowledge corresponding to the user query information is matched from the pre-constructed medical knowledge base; the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the medication query information. The technical solution of the present application can automatically recommend drugs for the medication query information input by the user, thereby improving the efficiency and accuracy of drug recommendation, and after predicting the candidate recommended drugs, the candidate recommended drugs can be further verified in combination with the drug information of the candidate recommended drugs and the medical knowledge obtained from the medical knowledge base that matches the user query information, and the final drug recommendation result is determined, so that the accuracy of drug recommendation is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0035] Figure 1 A schematic diagram of a drug recommendation method provided in an embodiment of the present application;

[0036] Figure 2 A schematic diagram of a process for determining a drug recommendation result provided in an embodiment of the present application;

[0037] Figure 3 A schematic diagram of a process flow for predicting candidate recommended drugs provided in an embodiment of the present application;

[0038] Figure 4 A schematic diagram of the structure of a drug recommendation device provided in an embodiment of the present application;

[0039] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] The technical solution of the embodiment of the present application is applicable to the application scenario of drug recommendation. The technical solution of the embodiment of the present application can reduce labor costs and improve the efficiency and accuracy of drug recommendation.

[0041] In the field of healthcare, drug recommendation is an important link. Existing drug recommendation methods usually rely on doctors' professional knowledge and experience, and doctors recommend drugs to patients. However, in most cases, face-to-face communication with doctors is required, which makes drug recommendation inefficient and doctors have a large workload. Long working hours may affect the accuracy of drug recommendations.

[0042] With the gradual accumulation of data, the development of big data and artificial intelligence technologies, and the improvement of people's health awareness, the intelligent drug recommendation system has begun to gain attention and gradually made progress. The role of the intelligent drug recommendation system is mainly to recommend drugs that meet the patient's conditions based on the patient's description requirements, provide help to patients, and save patients' time and energy. It is especially suitable for use in some occasions where it is unnecessary or inconvenient to contact a doctor. In addition, it can also be used for human-computer coupling to quickly provide doctors with treatment suggestions and reduce the workload of doctors. The system can be used alone or embedded in a larger intelligent medical system as a module. For example, after the active consultation of the human-computer dialogue, the system can be called based on the key information summarized during the consultation process to obtain the recommended drug results.

[0043] The existing intelligent drug recommendation system is a traditional drug recommendation system based on knowledge base and rules. This recommendation system mainly relies on a large amount of pre-written medical background knowledge, drug knowledge, judgment process and rules, etc. When the disease or symptoms described by the patient meet certain conditions, the corresponding drug will be recommended. However, this drug recommendation system relies on a large number of experts in the medical field to write relevant rules, which has high labor costs and is difficult to handle undefined new situations. For complex and changeable medical scenarios in reality, the accuracy of drug recommendations is low.

[0044] Based on this, the present application proposes a drug recommendation method, which can automatically recommend drugs for the medication inquiry information input by the user, further verify the candidate recommended drugs by combining the drug information of the candidate recommended drugs and the medical knowledge obtained from the medical knowledge base that matches the user's inquiry information, and determine the final drug recommendation results, thereby solving the problems of high labor costs and low efficiency and accuracy of drug recommendations in the prior art.

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] Exemplary Methods

[0047] See also Figure 1 As shown, the present application embodiment proposes a drug recommendation method. The method includes:

[0048] S101. Based on the user query information input by the user and / or the key information corresponding to the user query information, predict the candidate recommended drugs corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base.

[0049] When a user wants to inquire about the medicines needed for diseases and symptoms, the user needs to describe the diseases and symptoms he or she has when asking, that is, the user needs to enter user inquiry information containing the diseases and symptoms described by the user, for example, "My child has a cold, fever, and runny nose. What medicine should he or she take?", where the disease described by the user is "cold" and the symptoms are "fever, runny nose". In addition, the user inquiry information entered by the user may also include: the patient's special identity (for example, children, pregnant women, the elderly, etc.), the type of medicine the patient wants to use (for example, Chinese patent medicine, anti-inflammatory drugs, antibiotics, etc.), the method of medication the patient wants to use (for example, oral, ointment, spray, etc.), etc. In addition, in some complex scenarios, there may also be descriptions of the patient's own medication history or medication contraindications. In this embodiment, the user inputs text or voice through the user terminal, and the input text or voice is recognized to obtain user query information input by the user. If the user inputs text, the text is directly used as the user query information input by the user. If the user inputs voice, the existing voice-to-text algorithm is used to convert the user input voice into text, and the converted text is used as the user query information input by the user.

[0050] After obtaining the user inquiry information input by the user, this embodiment needs to extract key information from the user inquiry information to obtain the key information corresponding to the user inquiry information. If the user inquiry information contains the diseases and symptoms that the patient has, then the diseases and symptoms need to be extracted as key information. If the user inquiry information contains the patient's possible special identity, the method of medication that the patient wants to use, the type of medicine that the patient wants to use, etc., this information can also be extracted as key information. In this embodiment, the key information extraction of the user inquiry information can be achieved by using the element extraction method of the pre-trained deep learning model. This embodiment can use deep learning models of different scales to complete the extraction task according to different requirements for effect and efficiency in practical applications. For example, if the actual application has a high requirement for effect and a low requirement for efficiency, then a deep learning model with a larger scale can be selected to extract key information, which can make the key information highly accurate, but the key information extraction efficiency of the deep learning model is low. If the actual application does not have a high requirement for effect but a high requirement for efficiency, then a deep learning model with a smaller scale can be selected to extract key information, which can make the key information extraction efficiency of the deep learning model high, but due to the small size of the model, the accuracy of the extraction task is low. Therefore, this embodiment can construct and train a deep learning model according to the requirements for the efficiency and effect of key information extraction.

[0051] This embodiment preferably uses the factor extraction method of some small-scale pre-trained deep learning models (such as BERT, SpanBERT, etc.), directly connects the linear layer to the original output layer of the small-scale pre-trained language model, and then fine-tunes it to perform token-level classification, so as to extract the key information corresponding to the user's query information. If you want a better extraction effect and don't care too much about the cost, you can use the existing large language model to implement the extraction operation. These large models have powerful text comprehension capabilities because they have massive parameters and have been trained with massive data. You can specify them to extract key information from the text by writing instructions, and output it in a specific JSON format for easy processing. And this process generally does not require additional fine-tuning of the model, and you can use the existing public large models to simply call their interfaces to get the return results.

[0052] After extracting the corresponding key information from the user's query information, this embodiment can also standardize the key information according to a predefined vocabulary correspondence table and convert it into a unified medical term. For example, "cold" may be converted into "acute upper respiratory tract infection" and "fever" may be converted into "fever".

[0053] After obtaining the user query information and the key information corresponding to it, this embodiment predicts the candidate recommended drugs corresponding to the user query information by analyzing the user query information and / or the key information corresponding to the user query information. The user query information and / or the key information corresponding to the user query information are analyzed, and the user query information and / or the key information corresponding to the user query information can be matched with the data in the knowledge base containing medical medication knowledge, thereby querying the medication knowledge matching the user query information and / or the key information, and then taking the drugs recommended in the medication knowledge as candidate recommended drugs.

[0054] In addition, the present embodiment can also predict candidate recommended drugs through a pre-constructed large language model. Specifically, first construct a drug recommendation instruction to guide the large language model to predict candidate recommended drugs, and then input the user inquiry information and / or the key information corresponding to the user inquiry information, and the drug recommendation instruction into the large language model. The large language model recommends drugs according to the drug recommendation instruction and outputs the predicted candidate recommended drugs. The drug recommendation instruction can also set requirement information such as the number of drugs recommended by the large language model. For example, the constructed drug recommendation instruction can be: You are an experienced general practitioner who is familiar with various drug knowledge. Please recommend drugs according to the patient's requirements"( User requests for information and / or Key Information )" recommends appropriate drugs to patients and directly outputs ( Number of drugs ) generic names of drugs suitable for use, one for each line, no other content is needed. If the user's inquiry is "My child has a cold, fever, runny nose, what medicine should he take?", the number of drugs recommended by the preset large language model is 5. At this time, the corresponding drug recommendation instruction is: You are an experienced general practitioner who is familiar with various drug knowledge. Please recommend suitable drugs to the patient according to the patient's request "My child has a cold, fever, runny nose, what medicine should he take?", and directly output 5 generic names of drugs suitable for use, one for each line, and no other content is needed.

[0055] If the user query information and the corresponding drug recommendation instructions in the above example are input into the large language model, the output of the large language model may be as follows:

[0056] 1. Acetaminophen tablets

[0057] 2. Children's Paracetamol and Phenylephrine Granules

[0058] 3. Ibuprofen suspension

[0059] 4. Pseudoephedrine drops

[0060] 5. Loratadine syrup

[0061] The five drugs output by the above large language model are all used as candidate recommended drugs.

[0062] In order to improve the drug recommendation capability of the large language model and optimize the quality of the candidate recommended drugs, this embodiment can optimize the constructed drug recommendation instructions, wherein the optimization of the drug recommendation instructions includes but is not limited to the following methods:

[0063] First, use more auxiliary judgments to construct drug recommendation instructions. For example, in addition to information such as the patient's disease and symptoms, you can also add information such as the patient's special identity, the patient's desired method of medication, and the patient's desired type of drug. Use this information to build more complete drug recommendation instructions, so that the large language model can better meet user requirements when making drug recommendations, and the candidate recommended drugs obtained are more accurate.

[0064] Second, use the thinking chain method to construct drug recommendation instructions. Before outputting drugs, think about the patient's problem step by step according to the set requirements, output the analysis results, and finally give the drugs. For example, you can first ask to output the type of drugs that can be used for a certain symptom, and then give commonly used drugs based on the drug type.

[0065] Third, a self-reflective approach is used to construct drug recommendation instructions. After outputting candidate recommended drugs for the first time, the large language model is required to output the indications of the candidate recommended drugs. Based on whether the indications of the output candidate recommended drugs are consistent with the user's query information and / or key information, some erroneous drug names caused by hallucinations of the large language model can be eliminated.

[0066] In addition, this embodiment also pre-constructs a drug knowledge base, which stores relevant knowledge of various drugs, wherein the relevant knowledge of drugs includes various information such as drug name, drug type, indication, usage and dosage, contraindications, precautions, etc. The drug knowledge base can be constructed by obtaining relevant information of various drugs from the "Pharmacopoeia of the People's Republic of China", or by parsing the instructions of drugs to obtain relevant information of various drugs.

[0067] After predicting the candidate recommended drugs corresponding to the user's query information through the above steps, this embodiment needs to query the relevant knowledge corresponding to each candidate recommended drug from the drug knowledge base based on the predicted candidate recommended drugs as the drug information of the candidate recommended drugs. Among them, querying the relevant knowledge corresponding to each candidate recommended drug from the drug knowledge base can adopt a text matching method to match the candidate recommended drugs with each drug name in the drug knowledge base, query the drug name matching the candidate recommended drug, and then extract the relevant knowledge corresponding to the drug name matching the candidate recommended drug in the drug knowledge base as the drug information of the candidate recommended drug.

[0068] S102: using the user query information and key information, match the medical knowledge corresponding to the user query information from a pre-built medical knowledge base.

[0069] This embodiment pre-builds a medical knowledge base, which includes medication knowledge and medical field knowledge obtained from various channels. This embodiment uses user query information and key information to match the medical knowledge corresponding to the user query information from the pre-built medical knowledge base.

[0070] Specifically, the medical knowledge base includes a medical expert opinion knowledge base and / or a medical field knowledge base, wherein the medical expert opinion knowledge base stores the medication opinions provided by medical experts for various diseases. The medical expert opinion knowledge base is compiled and reviewed by medical experts, and is more authoritative and practical. The medication opinions for diseases may include but are not limited to: descriptions of the most commonly used drugs for the disease, identification of suitable drugs for different symptoms, medication misunderstandings for commonly used drugs, rapid response opinions for recent epidemic diseases, etc. The medical field knowledge base stores medical field knowledge generated based on medical literature, clinical data, and online medical knowledge. Medical literature includes textbooks and medical guidelines for medical majors, etc. Online medical knowledge includes popular science articles in medical-related public accounts / Baijiahao / Zhihu columns, and questions and answers between patients and doctors on various medical Q&A websites, etc. The content in the medical field knowledge base may be less authoritative and accurate than the content in the medical expert opinion knowledge base, but the data volume is huge, the acquisition is simple, and the coverage is comprehensive.

[0071] Furthermore, when the medical knowledge base includes a medical expert opinion knowledge base, this embodiment uses the user query information and key information to match the medical knowledge corresponding to the user query information from the pre-built medical knowledge base, specifically including the following steps:

[0072] The key information is matched with the medication opinions in the medical expert opinion knowledge base, and the medication opinions matching the key information are used as the medical knowledge corresponding to the user's query information.

[0073] The construction of the medical expert opinion knowledge base is directly compiled and reviewed by medical experts. This embodiment directly matches the key information corresponding to the user's query information with each medication opinion in the medical expert opinion knowledge base, and queries the medication opinions that match the key information as the medical knowledge corresponding to the user's query information. Among them, the key information is matched with each medication opinion in the medical expert opinion knowledge base, and the degree of matching between the key information and each medication opinion can be calculated, and the medication opinion matched by the key information can be selected according to the degree of matching. For example, two of the three information in the key information match the medication opinion of A, and only one of the three information in the key information matches the medication opinion of B, then it means that the degree of matching between the key information and the medication opinion of A is higher than the degree of matching between the key information and the medication opinion of B.

[0074] Furthermore, when the medical knowledge base includes a medical field knowledge base, this embodiment uses the user query information and key information to match the medical knowledge corresponding to the user query information from the pre-built medical knowledge base, specifically including the following steps:

[0075] The user query information is matched with the medical field knowledge in the medical field knowledge base, and the medical field knowledge matched with the user query information is used as the medical knowledge corresponding to the user query information.

[0076] This embodiment matches the user query information with the medical field knowledge in the medical field knowledge base, calculates the matching degree between the user query information and each medical field knowledge, and uses the medical field knowledge with a higher matching degree as the medical knowledge corresponding to the user query information. In calculating the matching degree between the user query information and each medical field knowledge, firstly, the user query information and the medical field knowledge are encoded respectively to obtain corresponding vectors, and then the matching score between the vector of the user query information and the vector of the medical field knowledge is calculated (for example, by using a dot product calculation method), and the medical field knowledge with the top N matching scores is selected as the medical knowledge corresponding to the user query information according to the sorting method of the matching scores in descending order.

[0077] For example, the medical field knowledge in the medical field knowledge base can be a question-and-answer pair between a patient and a doctor in a medical question-and-answer website. The user query information is matched with the question in the question-and-answer pair. If the matching degree between the user query information and the question in the question-and-answer pair reaches a preset threshold (such as 80%), the answer in the question-and-answer pair is used as the medical knowledge corresponding to the user query information. If the medical field knowledge in the medical field knowledge base is not a question-and-answer pair, but only a medical field knowledge text, the text similarity between the user query information and the medical field knowledge text is calculated as the matching degree. If the matching degree reaches the preset threshold, the medical field knowledge is directly used as the medical knowledge corresponding to the user query information.

[0078] This embodiment can also construct and train a matching model, input user query information and medical field knowledge into the matching model, use the matching model to predict the similarity between the user query information and each medical field knowledge in the medical field knowledge base, and then select the medical field knowledge with higher similarity from all the medical field knowledge as the medical knowledge corresponding to the user query information.

[0079] This embodiment can also combine the two methods of vector calculation and matching model. First, the user query information and medical field knowledge are encoded respectively to obtain corresponding vectors, and the matching score between the vector of the user query information and the vector of the medical field knowledge is calculated. According to the matching score descending sorting method, the medical field knowledge with the top N matching scores is selected as the candidate medical knowledge; then the user query information and the candidate medical knowledge are input into the matching model, and the matching model is used to predict the similarity between the user query information and each candidate medical knowledge, and then the candidate medical knowledge with higher similarity is selected from all the candidate medical knowledge as the medical knowledge corresponding to the user query information.

[0080] In addition, the construction of the medical field knowledge base uses web crawler technology to obtain a large amount of web page data, cleans the obtained web page data and stores it in the medical field knowledge base. Before collecting network data, it is necessary to conduct random inspections on the data from various sources to determine whether the data quality of each source is qualified. For example, for a variety of medical question-and-answer websites, after sampling inspection, you can select websites with a relatively high accuracy rate of doctors' answers to collect information, and you can filter more reliable answers by the hospital level, department, and professional title of the doctor marked on the website. After sampling inspection, select a website with a relatively high accuracy rate of doctors' answers, and you can mark and score hundreds of pieces of information by random sampling, and give a comprehensive evaluation from various aspects such as whether there are medical errors and whether the expression is accurate. Only websites with a high enough excellence rate will be considered.

[0081] In addition, for the method of checking text content and medical knowledge, a large language model can also be used for checking. For example, a knowledge checking instruction is pre-built, and the knowledge to be checked and the knowledge checking instruction are input into a pre-trained large language model, and the large language model is used to review the knowledge to be checked to determine whether the knowledge to be checked is consistent with the authoritative knowledge. Among them, the pre-built knowledge checking instruction can be as follows:

[0082] You are a medical expert. Please analyze the given knowledge to be examined based on the given authoritative medical knowledge, and judge its relationship with the authoritative knowledge based on medical common sense.

[0083] Possible relationships between the two include:

[0084] Unable to judge: The authoritative knowledge does not mention the content of the knowledge to be checked, making it impossible to determine whether it is supportive or opposed.

[0085] Supported: The content of the knowledge to be checked is mentioned in authoritative knowledge and is consistent with it in combination with medical common sense.

[0086] Objection: The authoritative knowledge mentions the content of the knowledge to be checked, and combined with medical common sense, it is in obvious conflict with it.

[0087] S103: Integrate and infer the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information.

[0088] After determining the candidate recommended drugs, the drug information of the candidate recommended drugs, and the medical knowledge corresponding to the user's query information through the above steps, this embodiment needs to combine and analyze the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information to determine whether the candidate recommended drugs are accurate, delete inaccurate drugs among the candidate recommended drugs, and use the remaining recommended drugs in the final candidate recommended drugs as the drug recommendation results corresponding to the user's query information, where the drug recommendation results can also include relevant information such as drug information and medication opinions of the recommended drugs obtained from the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information.

[0089] Specifically, the drug information of the candidate recommended drugs obtained includes various information such as drug type, indications, usage and dosage, contraindications, precautions, etc. By analyzing whether the drug information of the candidate recommended drugs is consistent with the user's inquiry information, it is better to confirm whether the candidate recommended drugs are really suitable for recommendation. The medication opinions in medical knowledge are written and reviewed by medical experts, which are more authoritative and practical. By analyzing whether the medication opinions are consistent with the drug information of the candidate recommended drugs, it is better to confirm whether the candidate recommended drugs are really suitable for recommendation.

[0090] Furthermore, this embodiment can pre-construct and train a large language model, and construct knowledge integration reasoning instructions, and input user query information, drug information of candidate recommended drugs, medical knowledge corresponding to the user query information, and knowledge integration reasoning instructions into the large language model. The large language model can integrate and reason about the user query information, drug information of candidate recommended drugs, and medical knowledge corresponding to the user query information, delete inaccurate candidate recommended drugs, and output the final drug recommendation results.

[0091] At present, the large language model has not yet developed to a very perfect stage. It may happen that the candidate recommended drugs are not so accurate, the incorrect indications are given, or the most commonly used drugs are not successfully recommended. At this time, the medical experts' medication opinions included in the medical knowledge can correct such errors in time, avoid the final output of wrong results, and improve the accuracy and practicality of the output. Compared with the medical field knowledge included in the medical knowledge and the medication opinions of medical experts, this knowledge may be less authoritative and accurate, but it is huge in quantity, easy to obtain, and comprehensive in coverage, and can also play an auxiliary role.

[0092] From the above introduction, it can be seen that the drug recommendation method proposed in the embodiment of the present application predicts the candidate recommended drugs corresponding to the user query information based on the user query information input by the user and / or the key information corresponding to the user query information, and obtains the drug information of the candidate recommended drugs from the pre-constructed drug knowledge base; wherein, the key information is obtained by extracting the key information of the user query information; using the user query information and the key information, the medical knowledge corresponding to the user query information is matched from the pre-constructed medical knowledge base; the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the medication query information. The technical solution of this embodiment can automatically recommend drugs for the medication query information input by the user, improve the efficiency and accuracy of drug recommendation, and, after predicting the candidate recommended drugs, the candidate recommended drugs can be further verified in combination with the drug information of the candidate recommended drugs and the medical knowledge matched with the user query information obtained from the medical knowledge base, and the final drug recommendation result is determined, so that the accuracy of drug recommendation is further improved.

[0093] As an optional implementation, see Figure 2 As shown, in another embodiment of the present application, step S103 in the above embodiment, integrating and reasoning the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information, specifically includes the following steps:

[0094] S201. Based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information, integrate and reason the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information.

[0095] When integrating and reasoning the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information, if there is a conflict between the drug information and the medical knowledge, it is necessary to select more accurate content from them as the basis for drug recommendation. Therefore, in order to improve the accuracy of drug recommendation, this embodiment needs to prioritize the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information, and use information with higher authority and accuracy as the basis for drug recommendation. That is, based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information, the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information are integrated and reasoned to determine the drug recommendation result corresponding to the user's query information.

[0096] Specifically, the priority of the medical knowledge corresponding to the drug information of the candidate recommended drugs and the user query information is set as follows:

[0097] 1. Regarding medication opinions contained in the medical knowledge corresponding to the user's inquiry information: Since medication opinions are written and reviewed by medical experts, they are of the highest quality and generally more timely. They can be updated in a timely manner for unseen diseases or the latest drug development or ban results. They are the most trustworthy of several knowledge sources. Therefore, the priority of medication opinions contained in the medical knowledge corresponding to the user's inquiry information can be set to the highest, and any relevant opinions can be given priority.

[0098] 2. Regarding the medical field knowledge contained in the medical knowledge corresponding to the user's inquiry information: Since there are a large number of patient-doctor question-and-answer resources on the Internet, the coverage of this part of the data is the highest. However, many websites do not have a strict content review mechanism, which often leads to doctors' random, imprecise or even wrong answers, which may interfere with drug recommendations and output incorrect results. However, the data on the Internet is huge in quantity, easy to obtain, and comprehensive in coverage. Therefore, the priority of the medical field knowledge contained in the medical knowledge corresponding to the user's inquiry information can be set to the lowest, which mainly plays a bottom-line role.

[0099] 3. For candidate recommended drugs and drug information of candidate recommended drugs: Since the big language model predicting candidate recommended drugs has been trained with massive data, the drugs recommended by the big language model are more credible to a certain extent. Therefore, the priority of candidate recommended drugs and drug information of candidate recommended drugs is set to medium, which is lower than the medication opinions of medical experts, but higher than the medical field knowledge from the Internet.

[0100] Furthermore, this step specifically includes:

[0101] First, based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information, a knowledge integration reasoning instruction is constructed.

[0102] This embodiment constructs a knowledge integration reasoning instruction based on the respective priorities of the drug information of the predetermined candidate recommended drugs and the medical knowledge corresponding to the user query information. For example, the knowledge integration reasoning instruction can be set as follows:

[0103] Please refer to the "Background Knowledge" and "Task Requirements" below to answer the "User Question".

[0104] Requirements

[0105] 1. Please refer to the "Knowledge Summary" section in each "Background Knowledge", combine different knowledge sources and candidate drugs, answer the "User Question" section, and give appropriate drug recommendation answers.

[0106] 2. When there are disagreements among multiple sources of knowledge, high-priority knowledge should be given priority.

[0107] 3. The response needs to reflect humanistic care and avoid medical risks.

[0108] Background

[0109] [{"priority":"high","knowledge summary":"medication advice"},{"priority":"medium","knowledge summary":"drug information 1 of candidate recommended drugs"},{"priority":"medium","knowledge summary":"drug information 2 of candidate recommended drugs"},{"priority":"medium","knowledge summary":"drug information 3 of candidate recommended drugs"},{"priority":"low","knowledge summary":"medical field knowledge 1"},{"priority":"low","knowledge summary":"medical field knowledge 2"}]

[0110] User Questions

[0111] {This is the user query information entered by the user}

[0112] Second, the user query information, the drug information of the candidate recommended drugs, the medical knowledge corresponding to the user query information and the knowledge integration reasoning instructions are input into the pre-trained first language model to obtain the drug recommendation results corresponding to the user query information.

[0113] This embodiment can pre-construct and train a first large language model, and input user query information, drug information of candidate recommended drugs, medical knowledge corresponding to the user query information, and knowledge integration reasoning instructions into the pre-trained first large language model. The first large language model integrates and reasons the user query information, drug information of candidate recommended drugs, and medical knowledge corresponding to the user query information according to the knowledge integration reasoning instructions, summarizes various knowledge contents, eliminates possible erroneous parts in the candidate recommended drugs, and finally gives reliable drug recommendation results.

[0114] In this embodiment, the large language model itself is in a "black box" state, its output is basically uncontrollable, and the training cost is often very high, and even many large language models do not have an open training interface. In this case, the medical experts' medication opinions included in the medical knowledge can be given a higher knowledge priority, adding a way of manual intervention. In addition to the above-mentioned ability to "patch" some bad large language model outputs in a timely manner, this can quickly adapt to unseen diseases or the latest drug development or banned results without having to wait for the large language model to be iteratively updated, ensuring the timeliness of the recommendation results. In addition, combining the drug recommendations of the large language model, the medication opinions of medical experts, and the medical field knowledge retrieved from the Internet knowledge, by making drug recommendations from knowledge from different sources, the large language model recommendation ensures the accuracy of the information, the medication opinions of medical experts ensure the professionalism and timeliness of the information, and the medical field knowledge retrieved from the Internet knowledge ensures the comprehensiveness of the information. The three complement each other and ultimately give the best drug recommendation results.

[0115] As an optional implementation, see Figure 3 As shown, in another embodiment of the present application, it is disclosed that predicting the candidate recommended drugs corresponding to the user query information based on the user query information input by the user and / or the key information corresponding to the user query information in step S101 of the above embodiment specifically includes the following steps:

[0116] S301. Input the user query information input by the user and / or the key information corresponding to the user query information, as well as the pre-built drug recommendation instructions, into the pre-trained second largest language model to obtain candidate recommended drugs corresponding to the user query information.

[0117] This embodiment pre-constructs and trains the second largest language model, and pre-constructs drug recommendation instructions, and inputs the user query information input by the user and / or the key information corresponding to the user query information, as well as the drug recommendation instructions into the second largest language model, so that the second largest language model performs drug recommendations and obtains candidate recommended drugs corresponding to the user query information.

[0118] S302. Input the user query information, the drug information of the candidate recommended drugs and the pre-built matching instructions into the second largest language model, so that the second largest language model determines whether the candidate recommended drugs are consistent with the user query information according to the matching instructions, and deletes the candidate recommended drugs that do not match the user query information from all candidate recommended drugs.

[0119] After the second largest language model predicts the candidate recommended drugs, the drug information of the candidate recommended drugs can be queried from the pre-constructed drug knowledge base. And the matching instructions are constructed again, and the user query information, the drug information of the candidate recommended drugs and the pre-constructed matching instructions are input into the second largest language model. The second largest language model determines whether the drug information of the candidate recommended drugs is consistent with the user query information according to the matching instructions. If the drug information of the candidate recommended drugs does not match the user query information, it means that the candidate recommended drugs output by the second largest language model are wrong, and the candidate recommended drugs are deleted from all candidate recommended drugs. The second largest language model also reflects on itself based on the result, thereby improving the accuracy of drug recommendations of the second largest language model. For example, if the user's query information requires cold medicine, and the indication in the drug information of the candidate recommended drug is diarrhea, the drug will be deleted from all candidate recommended drugs.

[0120] Exemplary Devices

[0121] Accordingly, the present application also provides a drug recommendation device, see Figure 4 As shown, the device comprises:

[0122] The drug prediction module 100 is used to predict the candidate recommended drugs corresponding to the user query information based on the user query information and / or the key information corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information;

[0123] A matching module 110 is used to match the medical knowledge corresponding to the user query information from a pre-built medical knowledge base using the user query information and key information;

[0124] The integrated reasoning module 120 is used to perform integrated reasoning on the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information, and determine the drug recommendation result corresponding to the user query information.

[0125] From the above introduction, it can be seen that the drug recommendation device proposed in the embodiment of the present application can automatically recommend drugs for the medication inquiry information input by the user, thereby improving the efficiency and accuracy of drug recommendations. Moreover, after predicting the candidate recommended drugs, the candidate recommended drugs can be further verified by combining the drug information of the candidate recommended drugs and the medical knowledge obtained from the medical knowledge base that matches the user's inquiry information, and the final drug recommendation results are determined, thereby further improving the accuracy of drug recommendations.

[0126] As an optional implementation, another embodiment of the present application discloses that the medical knowledge base includes a medical expert opinion knowledge base, which stores medication opinions provided by medical experts for various diseases; the matching module 110 is specifically used to:

[0127] The key information is matched with the medication opinions in the medical expert opinion knowledge base, and the medication opinions matching the key information are used as the medical knowledge corresponding to the user's query information.

[0128] As an optional implementation, another embodiment of the present application discloses that the medical knowledge base includes a medical field knowledge base, which stores medical field knowledge generated based on medical literature, clinical data, and network medical data; the matching module 110 is further specifically used for:

[0129] The user query information is matched with the medical field knowledge in the medical field knowledge base, and the medical field knowledge matched with the user query information is used as the medical knowledge corresponding to the user query information.

[0130] As an optional implementation, another embodiment of the present application discloses that the integrated reasoning module 120 is specifically used for:

[0131] Based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information, the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user's query information are integrated and reasoned to determine the drug recommendation result corresponding to the user's query information.

[0132] As an optional implementation, another embodiment of the present application discloses that the integrated reasoning module 120 includes: an instruction building unit and a model processing unit.

[0133] An instruction construction unit, used for constructing a knowledge integration reasoning instruction based on the priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information;

[0134] The model processing unit is used to input user query information, drug information of candidate recommended drugs, medical knowledge corresponding to the user query information and knowledge integration reasoning instructions into the pre-trained first language model to obtain drug recommendation results corresponding to the user query information.

[0135] As an optional implementation, another embodiment of the present application discloses that the drug prediction module 100 includes: a model prediction unit.

[0136] The model prediction unit is used to input the user query information input by the user and / or the key information corresponding to the user query information, as well as the pre-built drug recommendation instructions, into the pre-trained second largest language model to obtain the candidate recommended drugs corresponding to the user query information.

[0137] As an optional implementation, another embodiment of the present application discloses that the drug prediction module 100 further includes: a model reflection unit.

[0138] The model reflection unit is used to input user query information, drug information of candidate recommended drugs and pre-built matching instructions into the second largest language model, so that the second largest language model determines whether the candidate recommended drugs are consistent with the user query information according to the matching instructions, and deletes the candidate recommended drugs that do not meet the user query information from all candidate recommended drugs.

[0139] The drug recommendation device provided in this embodiment belongs to the same application concept as the drug recommendation method provided in the above embodiments of this application, and can execute the drug recommendation method provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of executing the drug recommendation method. For technical details not fully described in this embodiment, please refer to the specific processing content of the drug recommendation method provided in the above embodiments of this application, and will not be repeated here.

[0140] Exemplary Electronic Devices

[0141] Another embodiment of the present application also provides an electronic device, see Figure 5 As shown, the device includes:

[0142] Memory 200 and processor 210;

[0143] The memory 200 is connected to the processor 210 and is used to store programs;

[0144] The processor 210 is used to implement the drug recommendation method disclosed in any of the above embodiments by running the program stored in the memory 200.

[0145] Specifically, the electronic device may further include: a bus, a communication interface 220 , an input device 230 and an output device 240 .

[0146] The processor 210, the memory 200, the communication interface 220, the input device 230 and the output device 240 are connected to each other via a bus.

[0147] A bus may include a pathway that transfers information between components of a computer system.

[0148] The processor 210 may 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 for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0149] The processor 210 may include a main processor, and may also include a baseband chip, a modem, and the like.

[0150] The memory 200 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 200 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0151] The input device 230 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0152] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0153] The communication interface 220 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0154] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement each step of any drug recommendation method provided in the above embodiments of the present application.

[0155] Exemplary computer program products and storage media

[0156] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the drug recommendation method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0157] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0158] In addition, an embodiment of the present application may also be a storage medium on which a computer program is stored, and the computer program is executed by a processor to execute the steps of the drug recommendation method according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0159] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0160] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0161] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.

[0162] The modules and sub-modules in the devices and terminals of the various embodiments of the present application can be combined, divided and deleted according to actual needs.

[0163] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0164] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0165] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.

[0166] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0167] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0168] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0169] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be 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 the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A drug recommendation method, characterized in that: include: Based on the user query information input by the user and / or the key information corresponding to the user query information, predict the candidate recommended drugs corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information; Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base; The drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the user query information.

2. The method according to claim 1, characterized in that The medical knowledge base includes a medical expert opinion knowledge base, which stores medication opinions provided by medical experts for various diseases; Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base includes: The key information is matched with the medication opinions in the medical expert opinion knowledge base, and the medication opinions matched with the key information are used as the medical knowledge corresponding to the user query information.

3. The method according to claim 1, characterized in that The medical knowledge base includes a medical field knowledge base, which stores medical field knowledge generated based on medical literature, clinical data, and network medical data; Using the user query information and the key information, matching the medical knowledge corresponding to the user query information from a pre-built medical knowledge base includes: The user query information is matched with the medical field knowledge in the medical field knowledge base, and the medical field knowledge matched with the user query information is used as the medical knowledge corresponding to the user query information.

4. The method according to claim 1, characterized in that: Integrating and reasoning the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information includes: Based on the respective priorities of the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information, the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information are integrated and reasoned to determine the drug recommendation result corresponding to the user query information.

5. The method according to claim 4, characterized in that Based on the respective priorities of the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information, integrating and reasoning the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information, including: Constructing a knowledge integration reasoning instruction based on the priorities of the drug information of the candidate recommended drug and the medical knowledge corresponding to the user query information; The user query information, the drug information of the candidate recommended drugs, the medical knowledge corresponding to the user query information and the knowledge integration reasoning instructions are input into a pre-trained first large language model to obtain a drug recommendation result corresponding to the user query information.

6. The method according to claim 1, characterized in that Predicting a candidate recommended drug corresponding to the user query information based on the user query information input by the user and / or key information corresponding to the user query information includes: The user query information input by the user and / or the key information corresponding to the user query information, as well as the pre-constructed drug recommendation instructions, are input into the pre-trained second largest language model to obtain candidate recommended drugs corresponding to the user query information.

7. The method according to claim 6, characterized in that After inputting the user query information input by the user and / or the key information corresponding to the user query information, and the pre-built drug recommendation instruction into the pre-trained second language model to obtain the candidate recommended drugs corresponding to the user query information, the method further includes: The user query information, the drug information of the candidate recommended drugs and the pre-constructed matching instructions are input into the second largest language model, so that the second largest language model determines whether the candidate recommended drugs are consistent with the user query information according to the matching instructions, and deletes the candidate recommended drugs that do not match the user query information from all candidate recommended drugs.

8. A drug recommendation device, characterized in that: include: A drug prediction module, used to predict the candidate recommended drugs corresponding to the user query information based on the user query information input by the user and / or the key information corresponding to the user query information, and obtain the drug information of the candidate recommended drugs from a pre-built drug knowledge base; wherein the key information is obtained by extracting the key information of the user query information; A matching module, used to match the medical knowledge corresponding to the user query information from a pre-built medical knowledge base using the user query information and the key information; The integrated reasoning module is used to perform integrated reasoning on the drug information of the candidate recommended drugs and the medical knowledge corresponding to the user query information to determine the drug recommendation result corresponding to the user query information.

9. An electronic device, characterized in that: include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the drug recommendation method as described in any one of claims 1 to 7 by running the program in the memory.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, it implements the drug recommendation method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that The method comprises computer program instructions, which, when executed by a processor, enable the processor to implement the drug recommendation method as claimed in any one of claims 1 to 7.

Images