Medication question and answer management method, system and equipment based on intelligent agent

Abstract

The invention provides an agent-based medication question and answer management method, system and equipment, and the method comprises the steps: disassembling a demand according to demand basic information inputted by a user; based on a visual arrangement tool, creating a workflow which comprises an intention recognition node, a database interaction node and a large language model generation answer node; the method comprises the following steps of: standardizing a medicine specification and a user health file by utilizing a database tool, defining a workflow node data calling dynamic calling logic, preferentially calling database data, and triggering large model reasoning when the data is insufficient; large language model texts, voices and database pictures are utilized, and the intelligent agent API is connected with the interaction layer to output question and answer results. The answer is output in a question and answer mode, meanwhile, medicine consultation service, medicine reminding and medicine deadline management are provided for family users in combination with existing information of the database and large model reasoning, the situation that the medicine is still used after expiration is avoided, and the method has the advantages of being easy to operate, intelligent and convenient to use.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medication question and answer management, and particularly relates to a medication question and answer management method, system and device based on an intelligent agent. BACKGROUND

[0002] With the rapid development of medical informatization and artificial intelligence technology, the digitalization of medication management and consultation services is continuously improving. Traditional medication management mainly relies on doctor face-to-face consultation, pharmacist consultation and paper prescription management. Patients often face problems such as inconvenient information acquisition, poor medication compliance and chaotic drug management during medication. In recent years, various medication APPs and intelligent health management systems have emerged, trying to improve the current medication management situation through digital means.

[0003] Existing medication management systems usually adopt simple reminder functions and basic drug information queries, and some systems introduce artificial intelligence technology for medication guidance. However, these systems still have many defects and deficiencies in actual application: (1) Low degree of intelligence: existing systems mostly adopt a single question and answer mode, lack of specialized intelligent agents for different medication scenarios, and cannot provide accurate personalized services according to the specific consultation intentions of users, resulting in insufficient pertinence and professionalism of medication guidance.

[0004] (2) Limited multi-modal interaction capability: most systems only support text input, and have weak processing capabilities for voice consultation and image recognition (such as prescription sheets and drug packaging), resulting in poor user interaction experience, especially for the elderly user group, which has a high usage threshold.

[0005] (3) Single medication management function: existing systems mainly focus on medication reminder functions, lack comprehensive medication life cycle management, and cannot effectively integrate prescription recognition, medication plan development, missed dose handling, home drug inventory management and other links.

[0006] (4) Lack of family linkage mechanism: traditional medication management systems are centered on individuals, ignoring the important role of family members in medication management, especially for special groups such as the elderly and children, lacking effective family member coordination and supervision mechanisms.

[0007] (5) Insufficient emergency handling capability: existing systems lack intelligent identification and rapid response mechanisms when facing medication emergencies, and cannot identify help signals in time and provide corresponding emergency contact services.

[0008] (6) Low data utilization efficiency: the system cannot fully utilize users' historical medication data and health information, lacks long-term medication planning and chronic disease management capabilities based on personal health records, resulting in low service continuity and individualization.

[0009] (7) Lack of the function of double attributes of anti-feeding teaching. SUMMARY

[0010] The technical problem solved by the present application is to overcome the shortcomings of the prior art and provide a medication question and answer management method, system and device based on an agent, which provides medication consultation services and medication reminders, medication expiration management, and avoids the use of expired drugs.

[0011] The technical solution adopted by the present application to solve its technical problem is: a medication question and answer management method based on an agent, comprising the following steps: Requirement input disassembly, based on user input requirement basic information, and disassembling the requirement; Workflow modeling, based on a visual programming tool, sequentially creating a workflow including an intent recognition node, a database interaction node, and a large language model answer generation node; Multi-source data coupling, using a database tool to standardize the processing of drug instructions and user health records, and defining dynamic calling logic for workflow node data calling, and preferentially calling database data, and triggering large model reasoning when data is insufficient; Interactive output feedback, using large language model text, voice and database pictures, and connecting the interactive layer to output the question and answer results through the agent API.

[0012] Further, the requirement input disassembly comprises: Collecting user text, voice input or picture shooting through the API connection interactive layer to obtain user requirement basic information; Using natural language processing technology of a large language model node to analyze user requirement basic information and split user requirements.

[0013] Further, the workflow modeling comprises: Identifying user intent according to split user requirements, and dividing the identified user intent into nodes; Connecting the corresponding agent according to the intent node division, and searching for answers in the database corresponding to the agent; Generating an answer node using a large language model, and configuring the searched answer in the answer node.

[0014] Further, the agent comprises at least an intelligent health service body, an intelligent drug query body, and an intelligent drug reminder body.

[0015] Further, the database interaction node is set as a SQL generation rule and is associated with a drug database index.

[0016] Further, the large model reasoning adopts a bean bag large model.

[0017] Further, the API connection interaction layer at least includes a webpage, an APP and WeChat.

[0018] In a second aspect, the present application further discloses a medicine question and answer management system based on an agent, which adopts the medicine question and answer management method based on an agent, and comprises: An information input module, which is used for inputting the demand of a user and decomposing the demand of the user; A node calling module, which is connected with the information input module and calls corresponding agents according to the decomposed demand of the user; An agent module, which selects corresponding databases and large models for answer retrieval according to the decomposed demand of the user; An output feedback module, which utilizes large language model texts, voices and database pictures and outputs answers through the API connection interaction layer of the agent module.

[0019] Further, the agent module at least includes an intelligent health service body, an intelligent medicine query body and an intelligent medicine reminding body.

[0020] In a third aspect, the present application further discloses a medicine question and answer management device based on an agent, which adopts the medicine question and answer management method based on an agent, and comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the medicine question and answer management method when executing the program.

[0021] The present application has the following advantages: The present application adopts a question and answer mode to output answers, combines existing information in databases and large model reasoning, provides medicine consultation services and medicine reminders for family users, manages the period of medicines, and avoids the use of expired medicines. Secondly, the present application can be used for student teaching to improve the practical ability of students to review prescriptions, and has the characteristics of simple operation and intelligent convenience. DETAILED DESCRIPTION

[0022] Figure 1 A flowchart of the medicine question and answer management method based on an agent of the present application; Figure 2 A text input result display diagram of the medicine question and answer management method based on an agent of the present application; Figure 3 A picture input result display diagram of the medicine question and answer management method based on an agent of the present application; Figure 4 A first style structural schematic diagram of the medicine question and answer management device based on an agent of the present application; Figure 5This is a schematic diagram of the second version of the drug use question-and-answer management device based on intelligent agents according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1

[0025] like Figures 1 to 3 As shown, this invention discloses a medication question-and-answer management method based on intelligent agents, comprising the following steps: S100, Decompose the input requirements: Based on the basic information of the user's input requirements, decompose the requirements.

[0026] Specifically, the breakdown of the requirement input includes: (1) Use the API to connect to the interaction layer to collect user text, voice input or take pictures to obtain basic information about user needs.

[0027] The intelligent agent is defined by roles to regulate its specific service scope and skills, and completes drug search, medication reminders and management, and intelligent health services through three workflows. The drug search workflow displays the input content as images, voice, or text through a selector to facilitate different processing methods. If it is text, it directly inputs the large language model "Doubao-pro-32k / 241215" and sets its prompt words.

[0028] If the message is audio, it is converted to text using the "SPEECHTOTEXT" plugin. This is done by inputting the audio URL in string format as a parameter, which then calls the "Volcano Engine" model. As a speech recognition plugin for Volcano Engine, it supports text recognition from video links and files. The text is then input into a large language model to summarize the speech. "The user message contains audio; you need to summarize the content based on the audio content and the user's original message."

[0029] If it is a picture, the picture content is identified by the "IMGUNDERSTAND" plug-in. Picture understanding is a plug-in that generates meaningful and relevant text descriptions by analyzing the content of pictures on a specific URL. The plug-in uses deep learning algorithms to identify the main elements and activities in the picture and analyze the main content of the image to generate a text description related to it. In addition, the background of the picture can be analyzed to provide a text description of the shooting environment, such as weather, location, etc. Advanced machine vision and natural language processing techniques are used to help users understand the main content of the picture. The picture content is identified and summarized by a large language model, and the drug name and drug-related information are reorganized.

[0030] (2) Analyze the user's basic information and split the user's requirements based on the natural language processing technology of the large language model node.

[0031] The output value after splicing is input into the intent recognition node, and the summarized content is matched with the intent. The intents are "drug query", "children's drug query", and "disease popularization". For the drug query part, the stored drug name knowledge base is called, the large language model node is connected, the system prompt word is provided, and the output is input into the next large language model node to execute the automatic identification of drug key elements and search for drug information through Bing search function, and strictly follow the template output.

[0032] S200, workflow modeling, based on visual programming tools, create workflow including intent recognition node, database interaction node, large language model answer generation node.

[0033] The original voice and picture are converted into Chinese text through the "IMGUNDERSTAND" plug-in and "SPEECHTOTEXT" plug-in. Then the text is tokenized (Tokenization) by the tokenizer provided by the model, which is split into smaller word pieces. Through word piece to ID mapping, word embedding, and position encoding, the text is finally converted into a high-dimensional numerical vector sequence containing semantic information and position information. The sequence is input into the large language model node, and the code node is used to optimize the medication reminder time.

[0034] In the Medication_Alert workflow, the medication time before and after meals is dynamically optimized based on the standard time in the drug instruction and the past on-time medication rate. ; ; ; Alpha is the time deviation sensitive coefficient [0,1], different coefficients are taken for different drug key degrees, for example, insulin = 0.95, heart disease drug = 1, and vitamin = 0.3; Beta is the patient forgetfulness coefficient [0,1], which adjusts the reminder according to the user's habitual forgetting. Forgetful person = 0.9; meticulous person = 0.2; P is the sitting posture state monitoring {0,1}, which judges whether the user is in a sitting posture and whether it is suitable for taking medicine environment through external devices. Sitting posture = 1, non-sitting posture = 0; It is a real-time taking medicine state coefficient. If it is found through the reminder dialogue that this time is not taken, it is 1, otherwise it is 0; Theoretical taking medicine time, through drug instruction manual or user specification; Gamma is the attenuation adjustment factor (value 0.85-1), based on the recent compliance rate penalty coefficient, dynamically adjust the punishment strength, C is the number of times on time in the past 7 days; It is the repeated reminder time interval (which can be dynamically adjusted, the default value is 30 minutes); Unmedicated indicator, Mt = 1 → activated Item In the Medication_Alert workflow, it is connected to the output node through the code node.

[0035] In addition, there are Drug_inquiry, Intelligent_health_service workflows, and the workflow modeling includes: identifying the user's intention according to the split user demand, and dividing the identified user intention into nodes. According to the intention node division, connect the corresponding agent, and search for answers in the database corresponding to the agent. Use a large language model to generate an answer node, and configure the searched answer in the answer node. Among them, the agent at least includes an intelligent health service body, an intelligent drug query body, and an intelligent drug reminder body. In addition, the database interaction node is set to SQL generation rules, and is associated with a drug database index.

[0036] S300, multi-source data coupling, using database tools, standardizing the processing of drug instruction manuals and user health archives, and defining workflow node data calling dynamic calling logic, and preferentially calling database data, and triggering large model reasoning when data is insufficient.

[0037] Specifically, multi-source data coupling uses database tools to standardize data by standardizing the naming fields of drug instructions (in PDF format) and user health records (structured forms). In addition, workflow node data calling dynamic calling logic is defined, such as the drug query node prioritizing drug database calls and triggering large model reasoning when data is insufficient. Among them, the large model reasoning uses the bean bag large model.

[0038]

[0039] S400, interactive output feedback, using large language model text, voice and database pictures, converting text into a high-dimensional numerical vector sequence containing semantic information and location information, inputting the sequence into the large language model node, and outputting the question and answer results through the interaction layer (webpage, APP and WeChat) connected by the agent API.

[0040] In actual application, the user inputs the basic information "ankle pain" through "text", analyzes the input content provided by the user as "text" through the "Intelligent_health_service (Intelligent health service body)" workflow built by calling, and provides "symptoms" for the user through intent recognition and large model analysis, but does not know whether he has the disease, so the bean bag 1.5 pro is called, and the prompt word is fixed, combined with the preposed "text" basic information such as "ankle pain", "30 years old" and other basic information to fix the input format and content, and stable analysis, and through the next node to provide suggestions or health explanation. Input the picture of "Votalin" as the basic information, recognize the picture through the "Drug_inquiry (Intelligent drug query body)" workflow built by calling, and call the large model, fix the large model user prompt word and system prompt word, and stably output the drug information. Through the large model, the output format is fixed according to the template, and the output is "Mr. / Ms., the picture shows Votalin emulsion, which is a non-steroidal anti-inflammatory drug, mainly used for relieving mild to moderate pain such as joint pain, muscle pain, neuralgia, headache, etc. It can reduce pain and inflammation by inhibiting the synthesis and release of inflammatory mediators in the body. When using, it should be used according to the doctor's or drug instruction, and avoid overuse or long-term use". When I input the medication plan of paracetamol and warfarin, since the workflow "Medication_Alert (Intelligent drug reminder body)" calls the "drug interaction" database content through the large model, it triggers the "drug conflict" reminder and does not arrange the medication schedule.

[0041] At present, the intelligent agent is for family users and students. When the intelligent agent is for teaching, the intelligent agent can simulate a patient inquiry scene, students can have a dialogue with the intelligent agent (patient), train the students' pharmaceutical service ability, and the intelligent agent can make suggestions and evaluations on the previous dialogue. Second, I will establish an electronic prescription case library in the intelligent agent, and students can extract electronic prescriptions to train students' prescription auditing ability, automatically generate students' mistake sets, and generate result analysis.

[0042] Embodiment 2

[0043] The application discloses an intelligent agent-based medication question and answer management system, adopts the intelligent agent-based medication question and answer management method in embodiment 1, and comprises: An information input module is used for inputting the needs of users and disassembling the needs of users.

[0044] A node calling module is connected with the information input module, and the node calling module calls corresponding intelligent agents according to the disassembled needs of users. The intelligent agent module at least comprises an intelligent health service body, an intelligent drug query body and an intelligent drug reminding body.

[0045] An intelligent agent module is used for selecting corresponding databases and large models for answer retrieval according to the needs of users.

[0046] An output feedback module is used for outputting answers by using large language model texts, voices and database pictures and through the API connection of the intelligent agent module and the interaction layer.

[0047] Embodiment 3

[0048] The application discloses an intelligent agent-based medication question and answer management device, adopts the intelligent agent-based medication question and answer management method in embodiment 1, and comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the above-mentioned medication question and answer management method when the program is executed.

[0049] As shown in Figure 4 The device also has a display screen and function buttons. In the display screen area, the display screen displays expressions when the user has a conversation with the device, indicating that the current is in a conversation state. In the function button area, the user can perform power on / off operation, volume adjustment operation and remote call to friends and the like. At the same time, friends can send voice information to the device remotely, and the device can broadcast information to the user and the like.

[0050] As shown in Figure 5Another type of the device is shown. The device has a function button. When the button is pressed for a long time, the on-off operation is performed, when the button is pressed for a short time, the remote call is performed, the call information can be sent to the preset relatives and friends, and the online call can be performed. When the button is rotated, the volume can be adjusted.

[0051] Compared with the prior art, the present application has at least the following beneficial effects: The present application adopts the question and answer mode to output answers, combines the existing information in the database and the large model reasoning, provides the medication consultation service and the medication reminder for the family users, the drug period management, and avoids the use of the expired drug. Secondly, it can be used for student teaching, improves the practical ability of students to review the prescription, has the characteristics of simple operation and intelligent convenience.

[0052] The above technical features can be understood and implemented by the skilled in the art through the textual description, and therefore will not be described in the drawings.

[0053] Finally, it should be noted that: the above only describes the preferred embodiments of the present application, and is used to explain the technical scheme of the present application, but not to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application are included in the protection scope of the present application.

Claims

1. A medication question-and-answer management method based on intelligent agents, characterized in that, Includes the following steps: Decompose the input requirements: Based on the basic information of the user's input requirements, break down the requirements. Workflow modeling, based on a visual orchestration tool, sequentially creates workflows including intent recognition nodes, database interaction nodes, and large language model-generated response nodes; Multi-source data coupling utilizes database tools to standardize drug instructions and user health records, defines dynamic call logic for workflow node data calls, prioritizes database data calls, and triggers large model inference when data is insufficient. The interactive output feedback utilizes large language model text, speech, and database images, and outputs question-and-answer results through the intelligent agent API connected to the interaction layer.

2. The medication question-and-answer management method based on intelligent agents as described in claim 1, characterized in that, The breakdown of the demand input includes: Use the API to connect to the interaction layer to collect user text, voice input, or take pictures to obtain basic information about user needs; Natural language processing techniques based on large language model nodes are used to parse the basic information of user needs and break down the user's needs.

3. The medication question-and-answer management method based on intelligent agents as described in claim 1, characterized in that, The workflow modeling includes: Identify user intent based on the segmented user needs, and divide the identified user intent into nodes; The connection is divided into corresponding agents based on the intent node, and the answer is searched in the database corresponding to the agent. The large language model is used to generate answer nodes, and the search results are configured in these answer nodes.

4. The drug use question-and-answer management method based on intelligent agents as described in claim 3, characterized in that, The intelligent agent includes at least an intelligent health service agent, an intelligent drug query agent, and an intelligent drug reminder agent.

5. The medication question-and-answer management method based on intelligent agents as described in claim 1, characterized in that, The database interaction node is configured with SQL generation rules and associated with the drug database index.

6. The drug use question-and-answer management method based on intelligent agents as described in claim 1, characterized in that, The large-scale inference uses the Doubao large-scale model.

7. The drug use question-and-answer management method based on intelligent agents as described in claim 1, characterized in that, The API connection and interaction layer includes at least web pages, apps, and WeChat.

8. A drug question-and-answer management system based on intelligent agents, employing the drug question-and-answer management method based on intelligent agents as described in any one of claims 1-7, characterized in that, include: An information input module is used to input user needs and break down user needs; A node invocation module is connected to the information input module, and the node invocation module invokes the corresponding intelligent agent according to the decomposed user needs; The intelligent agent module selects the corresponding database and large model to retrieve answers based on the user's specific needs. The output feedback module utilizes large language model text, speech, and database images, and outputs the answer by connecting to the interaction layer through the intelligent agent module API.

9. The agent-based medication question-and-answer management system as described in claim 8, characterized in that, The intelligent agent module includes at least an intelligent health service agent, an intelligent drug query agent, and an intelligent drug reminder agent.

10. A drug use question-and-answer management device based on intelligent agents, employing the drug use question-and-answer management method based on intelligent agents as described in any one of claims 1-7, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the medication question-and-answer management method as described in any one of claims 1-7.

Images