Health education resource RPA interaction system and method based on knowledge graph

Abstract

The invention provides a health education resource RPA interaction system and method based on a knowledge graph, and the system comprises a multi-source data fusion module which is used for summarizing health data of medical literatures, electronic cases and monitoring equipment, calling a pre-training model for entity recognition and relation extraction, and constructing a dynamically updated health knowledge graph; the RPA process management module is used for generating an automatic process according to the health knowledge graph and allowing a user to dynamically adjust a generation rule; the multi-modal interaction module provides texts, languages and / or visual charts for the user for real-time interaction according to the generated automatic process; and the feedback optimization module is used for optimizing the knowledge graph and the RPA process according to the user behavior log. According to the invention, the RPA system is combined with the AI, the health problem of the patient is intelligently analyzed, task scheduling is automatically carried out, personalized and accurate health education service is provided, and the intelligence and practicability of health education are improved.

Description

Technical Field

[0001] The present application relates to the field of RPA intelligent interaction technology, and more specifically, to a health education resource RPA interaction system and method based on knowledge graph. Background Art

[0002] Health education is a crucial component of healthcare, aiming to improve patients' health knowledge and self-management abilities. In recent years, with the advancement of artificial intelligence (AI), health education has gradually become more intelligent. For example, technologies such as knowledge graphs, natural language processing, and machine learning have been widely applied to the development of health education platforms, significantly enhancing the relevance and interactivity of educational content. However, most of these platforms rely on manual operation and management, making efficient automation difficult.

[0003] Robotic process automation (RPA) is a software technology that simulates human processes to complete repetitive tasks. Combining RPA with AI technology can provide more flexible and intelligent solutions, demonstrating significant application potential in fields such as healthcare and finance. In particular, combining RPA with AI offers new possibilities for automated and personalized health education.

[0004] Among existing technologies, robotic systems based on the combination of artificial intelligence (AI) and robotic process automation (RPA) are gaining increasing attention. Many existing technologies leverage AI and RPA to provide intelligent services in various fields, interacting with users through natural language processing (NLP) and speech recognition.

[0005] For example, Chinese patent application CN202110024488.1 discloses a task scheduling method based on RPA and AI. This method uses an RPA robot to query task information from the task scheduling system, generate scheduling information, and convert the task content into voice and play it to the executor through a telephone robot. This technology can improve task scheduling efficiency and reduce manual intervention, and is particularly suitable for the automated processing of repetitive tasks. Although this solution has certain intelligent advantages in task scheduling, it mainly focuses on task scheduling and voice interaction, does not involve professional knowledge and personalized services in the field of health education, and cannot automatically perform complex operations related to user health needs.

[0006] Prior art example 2 Chinese patent application CN202210996195.4 describes a multi-dimensional, multifunctional intelligent robot for technical guidance in the operating room. This robot combines a medical artificial intelligence support system, an RPA robot platform, and a client. The AI-powered system analyzes and processes medical data, providing real-time surgical guidance. While this solution offers a high level of intelligence in medical guidance, it is limited to operating room applications and does not address the health education sector. In particular, it lacks adaptability to individual patient health needs and educational content.

[0007] Prior art example 3 Chinese patent application CN202411049619.1 provides a multi-agent AI robot-based supervision method for pharmaceutical products. This system uses a multi-agent architecture combined with a large language model and RPA robots to automate the supervision of drug production, distribution, and regulation. While this technology can improve supervision efficiency through multi-agent collaboration, it is primarily used in the pharmaceutical regulatory field and lacks the ability to respond to individualized needs in health education. It also fails to integrate RPA with AI for the generation and interaction of health education content.

[0008] Although these existing technical solutions have made certain progress in different application fields, they generally have problems such as the inability to effectively integrate medical expertise, the lack of interdisciplinary intelligent services, and insufficient response to personalized needs in health education. In addition, traditional RPA systems rely on fixed rules, cannot dynamically adapt to complex health scenarios, and lack a deep understanding of user questions. Therefore, the present invention proposes a health education resource RPA interaction system and method based on knowledge graphs, which can intelligently analyze patients' health problems, automatically schedule tasks, and provide personalized and accurate health education services, thereby addressing the limitations of existing technologies and improving the intelligence and practicality of health education. Summary of the Invention

[0009] In order to solve the above technical problems, this application is proposed. This application provides a health education resource RPA interaction system and method based on knowledge graph.

[0010] According to one aspect of the present application, a health education resource RPA interaction system based on a knowledge graph is provided, comprising: a multi-source data fusion module, an RPA process management module, a multimodal interaction module, and a feedback optimization module; The multi-source data fusion module is used to aggregate health data from medical literature, electronic medical records, and monitoring equipment, and uses pre-trained models for entity recognition and relationship extraction to build a dynamically updated health knowledge graph; The RPA process management module is used to generate automated processes based on the health knowledge graph and allows users to dynamically adjust the generation rules; The multimodal interaction module provides users with text, language and / or visual charts for real-time interaction based on the generated automated process; The feedback optimization module is used to optimize the knowledge graph and RPA process based on user behavior logs.

[0011] Preferably, the multi-source data fusion module cleans and standardizes the collected data, maps different terms for the same object in data from different sources into the same field, and unifies the units; uses the cleaned and standardized data to input the pre-trained model for entity recognition and relationship extraction, thereby constructing the required health knowledge graph.

[0012] Preferably, the RPA process management module generates an automated process based on the actual input content of the user and the constructed health knowledge graph, and sets an adjustment button in each node of the process to allow the user to adjust the order of the process nodes, add or delete a node process, and modify the specific content of a node process.

[0013] Preferably, the multimodal interaction module interacts with the user in real time, and the user queries and asks questions about the content displayed by the multimodal interaction module in real time through voice and text; and displays the latest output content.

[0014] Preferably, the feedback optimization module collects and analyzes the user's behavior logs during the use of the system, including the user's access frequency, residence time, interactive operation records, adjustment records of the generation process, and user feedback information for different contents of health education resources; generates satisfaction scores, optimizes and updates the health knowledge graph based on the satisfaction level, supplements and improves the entities and relationships in the knowledge graph; makes targeted adjustments to the RPA process, and optimizes the process steps and parameter settings.

[0015] The second aspect of the present invention provides a knowledge graph-based health education resource RPA interaction method, which is applied to the above-mentioned interactive system and includes the following specific steps: S1: collect data and preprocess the data; S2: Based on the preprocessed data, the pre-trained model is called to perform entity recognition and relationship extraction to build a dynamically updated health knowledge graph; S3: Generate an automated process based on the user's actual input and the constructed health knowledge graph; S4: providing users with text, language, and / or visual charts for real-time interaction based on the generated automated process; S5: Optimize the knowledge graph and RPA process based on user behavior logs.

[0016] Preferably, in step S3, semantic recognition is performed based on the actual input content of the user to determine the user's demand goals and potential goals; the goals are broken down into several subtasks that can be automated; and the subtasks are arranged in a logical order to output the final RPA process.

[0017] Preferably, it includes: analyzing the subtasks broken down from the demand targets and potential targets, and collaboratively processing the subtasks with the same processing steps or types.

[0018] The third aspect of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, it implements the steps of the above-mentioned knowledge graph-based health education resource RPA interaction method.

[0019] The fourth aspect of the present invention provides a computer-readable medium, which carries one or more programs. When the above one or more programs are executed by an electronic device, the electronic device implements the steps of the above-mentioned knowledge graph-based health education resource RPA interaction method.

[0020] Compared with the existing technology, the RPA interactive system and method for health education resources based on knowledge graphs provided in this application integrates various types of data through a multi-source data fusion module, adopts data cleaning standardization to achieve entity alignment across data sources, combines pre-trained models to complete entity recognition and relationship extraction, and constructs a health knowledge graph that supports dynamic node expansion; the graph-based RPA process management module generates customizable automated health management processes, provides a visual adjustment interface to support users to add, delete, modify and sort process nodes, and realizes two-way real-time interaction between natural language questions and answers and process diagrams through multimodal interaction; the feedback optimization module collects user behavior logs and satisfaction data, and adopts machine learning methods to continuously optimize the entity relationships and RPA process parameters of the knowledge graph, forming a closed-loop system of "data-driven-intelligent generation-interactive optimization", and ultimately achieving accurate recommendation and dynamic evolution of personalized health education. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a system block diagram of the knowledge graph-based health education resource RPA interaction system according to an embodiment of the present application.

[0022] Figure 2This is a flowchart of the RPA interaction method for health education resources based on knowledge graph according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0024] As mentioned in the above background technology, although there are some cases of combining AI and RPA, there are generally problems such as the inability to effectively integrate medical expertise, the lack of interdisciplinary intelligent services, and insufficient response to personalized needs in health education. In addition, traditional RPA systems rely on fixed rules and cannot dynamically adapt to complex health scenarios, and lack a deep understanding of user questions. In addition, traditional RPA systems require multiple rounds of questions and answers to obtain the user's real needs, which affects the user's actual experience. Therefore, the present invention proposes a health education resource RPA interaction system and method based on knowledge graphs, which can automatically perform task scheduling through intelligent analysis of patients' health problems, and provide personalized and accurate health education services, thereby solving the limitations of existing technologies and improving the intelligence and practicality of health education. Example 1

[0025] like Figure 1 As shown, this embodiment provides a health education resource RPA interaction system based on knowledge graph, including: a multi-source data fusion module, an RPA process management module, a multimodal interaction module and a feedback optimization module; In this embodiment, the multi-source data fusion module is used to aggregate the health data of medical literature, electronic medical records, and monitoring equipment, and call the pre-trained model for entity recognition and relationship extraction to construct a dynamically updated health knowledge graph; wherein, the multi-source data fusion module cleans and standardizes the collected data, maps different terms for the same object in data from different sources to the same field, and unifies the units; the cleaned and standardized data is used to input the pre-trained model for entity recognition and relationship extraction, thereby constructing the required health knowledge graph. For example, when integrating practical reference documents such as medical literature and clinical guidelines, the same object is often represented using different terms. For example, Alzheimer's disease (AD) requires mapping different terms for the same object to the same field and unifying the units used for medication. Different representations in different documents can be unified, for example, by using the JAPE algorithm to achieve entity alignment across data sources. Pre-trained models can then be used for entity recognition and relationship extraction to construct a dynamically updated health knowledge graph. The constructed health knowledge graph retains a node generation module that automatically expands the graph's nodes based on user feedback and the latest research data to achieve dynamic updates. For electronic medical records, user authorization is required, or the user manually uploads medical records. This allows for the combination of medical records to identify potential health issues or chronic diseases requiring long-term attention. Common health monitoring devices include smartwatches (which monitor the user's heart rate), blood glucose meters (which monitor the user's blood glucose data), and blood pressure monitors (which monitor the user's blood pressure data). User authorization is also required for monitoring data, and the user's real-time health status is determined based on the monitored health data.

[0026] In this embodiment, the RPA process management module is used to generate an automated process based on the health knowledge graph and allow users to dynamically adjust the generation rules; the RPA process management module generates an automated process based on the actual user input content and combines it with the constructed health knowledge graph, and sets an adjustment button in each node of the process, allowing users to adjust the order of process nodes, add or delete a node process, and modify the specific content of a node process. For example, when a user inputs the question "What dietary precautions should diabetic patients take" through the multimodal interaction module, the question content is combined with the constructed health knowledge graph to search and determine the related entities; the output automated process includes a correct diet list, a taboo diet list, optional medicines, dosages, and process nodes for side effects; the user can adjust the order of the output process nodes according to their preferences. For example, if the user only wants to prevent the risk of diabetes and does not want to take medication, the optional medicines, dosages, and process nodes for side effects can be directly deleted; the focus can be on the correct diet list and the taboo diet list; in the correct diet list and the taboo diet list, the user can also adjust the order of the foods in the above lists according to their preferences; after the process is determined, an executable automated process is generated and pushed to the user; the user can control his diet according to the process to achieve the purpose of adjusting his health and reducing the risk of disease.

[0027] In this embodiment, the multimodal interaction module provides users with text, language and / or visual charts for real-time interaction based on the generated automated process; the multimodal interaction module interacts with the user in real time, and the user queries and asks questions about the content displayed by the multimodal interaction module in real time through voice and text; the latest output content is displayed; in the multimodal interaction module, the user can choose to ask questions and have the system automatically process them and obtain the results of the questions; the form of the user's questions can be text input or voice input; the results of the questions output by the system can also be in text, language, or in the form of charts to facilitate user understanding; in the multimodal interaction module, the user can also ask questions again based on the output answers to the questions, and after obtaining a satisfactory answer to the user, the RPA process management module is called for the user to choose to adjust the corresponding automated process.

[0028] In this embodiment, the feedback optimization module is used to optimize the knowledge graph and RPA process based on user behavior logs. This module collects and analyzes user behavior logs during system usage, including frequency of access to different health education resource contents, dwell time, interaction records, adjustments to generated processes, and user feedback. It generates satisfaction scores and optimizes and updates the health knowledge graph based on these scores, supplementing and improving the entities and relationships within the knowledge graph. It also makes targeted adjustments to the RPA process, optimizing process steps and parameter settings.

[0029] This embodiment integrates medical literature, electronic medical records, and monitoring equipment data through a multi-source data fusion module, adopts data cleaning standardization to achieve entity alignment across data sources, combines pre-trained models to complete entity recognition and relationship extraction, and constructs a health knowledge graph that supports dynamic node expansion; the graph-based RPA process management module generates customizable automated health management processes and provides a visual adjustment interface to support users' operations of adding, deleting, modifying, and sorting process nodes; the multimodal interaction module integrates three interaction methods: text, voice, and visual charts, to achieve two-way real-time interaction between natural language questions and answers and process diagrams; the feedback optimization module collects user behavior logs and satisfaction data, and uses machine learning methods to continuously optimize the entity relationships and RPA process parameters of the knowledge graph, forming a closed-loop system of "data-driven-intelligent generation-interactive optimization", ultimately achieving accurate recommendations and dynamic evolution of personalized health education. Example 2

[0030] like Figure 2 As shown, this embodiment provides a health education resource RPA interaction method based on a knowledge graph, which is applied to the interactive system in Example 1 and includes the following specific steps; S1: Collect data and preprocess the data; first, use the multi-source data fusion module to collect data and perform preprocessing operations simultaneously; data sources can be divided into structured data, semi-structured data and unstructured data; common structured data include electronic medical records, laboratory test index databases, drug libraries, and standard clinical guidelines; semi-structured data include medical literature, clinical trial registries, and health device data; unstructured data include patient-reported symptoms and medical imaging reports; all of the above data need to be cleaned and preprocessed; for example, knowledge distillation technology can be used, with ChatGPT-4 as the teacher model, to correct grammatical errors in semi-structured data and unstructured data to improve the accuracy of the health knowledge graph finally constructed; it should be noted that the system can preprocess and standardize various types of health data to generate a general knowledge graph. After the user authorizes, it calls the user's electronic medical records and health device data, and generates new entities and relationships based on the general health knowledge graph, thereby generating a customized health knowledge graph for individual users; the newly generated entities and relationships are only open to the user himself.

[0031] The multi-source data fusion module cleans and standardizes the collected data, maps different terms for the same object in data from different sources to the same field, and unifies the units; the cleaned and standardized data is input into the pre-trained model for entity recognition and relationship extraction, thereby constructing the required health knowledge graph. For example, in the process of integrating medical literature, clinical guidelines and other reference documents with practical significance, it is often encountered that the same object is represented by different terms; for example, Alzheimer's disease is also known as AD, so it is necessary to uniformly map different terms of the same object to the same field in this step, and unify the units when taking medication; unify different expressions in different documents, for example, entity alignment across data sources can be achieved based on the JAPE algorithm; the entity alignment step has been disclosed in many documents in the prior art; such as the Chinese patent with publication number CN118780357A, a method for constructing a multi-source knowledge graph based on medical and health data, discloses the following technical solutions: obtain a medical and health data set; use a pre-trained knowledge distillation model to perform grammatical correction preprocessing on the medical and health data set; construct a schema based on the medical and health data after grammatical correction preprocessing; use the ERNIE-UIE model to extract knowledge from the constructed schema to obtain a single-source knowledge graph; perform entity alignment and attribute alignment on the single-source knowledge graph to obtain a multi-source knowledge graph. The detailed steps of the JAPE algorithm to achieve entity alignment across data sources are recorded in its specification: traditional entity alignment methods include using structural information for entity alignment, such as MTransE, BootEA, etc.; and the Schema-Agnostic algorithm that uses attribute information for entity alignment. When only structural information is used, the attribute information of the knowledge graph is ignored, and attribute information has the characteristics of higher accuracy and can improve the entity alignment effect. However, only using attribute information ignores the structure of the knowledge graph and the connection between entities. For this reason, this embodiment proposes the JAPE algorithm, an entity alignment method using structural information and attribute names. The basic idea of ​​the JAPE algorithm is to obtain the structural embedding SE based on the relationship triples, and obtain the attribute embedding EA based on the attribute triples, and finally SE and EA are combined. The embeddings of all entities in the two knowledge graphs are combined to achieve the purpose of entity alignment. The same scheme can be used for entity alignment in this embodiment, and other detailed steps are not repeated here.

[0032] S2: Based on the preprocessed data, the pre-trained model is called to perform entity recognition and relationship extraction to construct a dynamically updated health knowledge graph; then the pre-trained model can be used to perform entity recognition and relationship extraction to construct a dynamically updated health knowledge graph; the node generation module is retained in the constructed health knowledge graph to automatically expand the nodes of the graph according to user interaction feedback and the latest research data to achieve the purpose of dynamic updating.

[0033] For the pre-training module, commonly used adaptation models in the field, such as BioBERT and ClinicalBERT, can be used to identify medical entities; after labeling a small amount of data, high-confidence samples are screened through the model prediction results, and the model is iteratively optimized; in an optional embodiment, the text is input: "The patient is a male, 55 years old, fasting blood glucose 8.0, and the main complaint is dizziness", and the entity is output; entity 1: dizziness, type: symptom; entity 2: fasting blood glucose 8.0, type: test result; after the medical entity is identified, relationship extraction can be performed, and regular expressions can be matched for structured text, such as contraindications clearly stated in the drug instructions; possible relationships are established, for example: diabetes - use - metformin - applicable population - obese or overweight patients with type 2 diabetes - contraindicated - contraindicated in patients with renal insufficiency, severe liver disease or alcoholics; diabetes - use - acarbose - applicable population - patients with increased blood sugar after meals, especially those who mainly eat carbohydrates - contraindicated - patients with intestinal obstruction or severe renal insufficiency.

[0034] For the constructed health knowledge graph, evidence weighting and time decay are used to reduce existing contradictions or conflicts. For example, when the conclusion of a paper contradicts the patient's self-report, authoritative papers are given priority as the source. For example, the blood sugar control target recommended in the new treatment guidelines is used to replace the weight of outdated knowledge in the old guidelines. There are many documents on the method of constructing a knowledge graph. This embodiment only proposes a feasible solution to establish a knowledge graph. Other solutions can be used as steps for constructing a knowledge graph in this application. For example, the Chinese patent with publication number CN118780357A, a multi-source knowledge graph construction method based on medical and health data, can be used to construct the knowledge graph required for this solution.

[0035] In this embodiment, it should be noted that for electronic medical record data, user authorization is required, or the user manually uploads the medical record data, so as to facilitate the combination of the medical record data to determine the user's potential health problems or chronic diseases that require long-term attention; for health monitoring devices, common ones include smart watches (monitoring the user's heart rate), blood glucose meters (monitoring the user's blood glucose data), and blood pressure meters (monitoring the user's blood pressure data). For monitoring data, user authorization is also required to determine the user's real-time health status based on the monitored health data.

[0036] S3: Generate an automated process based on the actual user input and the constructed health knowledge graph; in step S3, perform semantic recognition based on the actual user input to determine the user's demand goals and potential goals; break down the goals into several subtasks that can be automated; arrange the subtasks in a logical order to output the final RPA process.

[0037] Specifically, after the user inputs a question through the multimodal interaction module, the system can call a large model interface with deep thinking capabilities, such as deepseek, kimi, etc.; use the existing large model to deeply think about the user's question, explore the user's possible needs, and sort them according to the degree of demand; analyze the subtasks decomposed from the demand goals and potential goals, and collaboratively process subtasks with the same processing steps or types.

[0038] The RPA process management module generates an automated process based on the actual user input and the constructed health knowledge graph. Adjustment buttons are set in each node of the process to allow users to adjust the order of process nodes, add or delete a node process, and modify the specific content of a node process. For example, when a user enters the question "What dietary precautions should diabetic patients take" through the multimodal interaction module, the module searches for the relevant entities based on the content of the question and the constructed health knowledge graph. When the above question is input into the large model for in-depth thinking, it is found that the user's surface demand is the diet that diabetic patients need to pay attention to, and the potential demand is the daily dietary habits of diabetic patients, and there are different points of attention for different patients. Type 1 diabetes is caused by an absolute lack of insulin, so the diet needs to be more precisely matched to the insulin dosage; Type 2 diabetes places more emphasis on low glycemic index (GI) foods and controlling total calories, especially for obese patients who need to lose weight; in addition, type 1 needs to pay special attention to hypoglycemia and carry sugary foods with them, while type 2 is more concerned about cardiovascular and kidney disease risks, so salt and saturated fat are limited in the diet. Therefore, it is necessary to combine the user's personal health information to output targeted automated processes for them; for example, for patients with type 1 diabetes, the output processes include: using the "carbohydrate counting method", with 1 unit of rapid-acting insulin for every 10 grams of carbohydrates, giving priority to low-GI foods (such as brown rice and oats), higher protein intake (1.0-1.5 grams / kg body weight), mainly lean meat and fish, and avoiding excessive red meat, Choose mono / polyunsaturated fatty acids (such as nuts and deep-sea fish) as fats, and trans fats need to be <10% of total calories. When fever or infection occurs, insulin requirements may increase by 50%-100%, and additional carbohydrates are needed. Blood sugar needs to be monitored after exercise to prevent delayed hypoglycemia. Low-GI fruits (such as apples and berries) can be eaten in moderation, but they need to be included in the total amount of carbohydrates and insulin should be adjusted. Sugary foods (such as bananas) should be chosen as snacks to deal with emergency hypoglycemia. For patients with type 2 diabetes, the output processes include: low GI staple foods account for more than 50% (such as oats and miscellaneous beans), daily total carbohydrates are controlled at 150-250 grams, vegetables are used to replace part of the staple food (such as 500 grams of vegetables ≈ 25 grams of rice and noodles), appropriate protein (1.0-1.2 grams / kg body weight), priority is given to plant protein (such as soy products), red meat is reduced, saturated fat (such as animal oil) is strictly limited, trans fat intake must be zero, obese patients need to lose 5%-10% weight through diet, reduce insulin resistance, avoid high-salt foods (≤5 grams per day), prevent hypertension and kidney disease, limit fruit to 200 grams per day, avoid high-sugar fruits (such as mangoes), and recommend eating between meals and choosing low-fat, high-fiber foods (such as nuts, ≤10 grams per day) as snacks.

[0039] In this embodiment, the output automated process includes a correct diet list, a taboo diet list, optional medicines, dosages, and process nodes for side effects; the user can adjust the order of the output process nodes according to preference. For example, if the user only wants to prevent the risk of diabetes and does not want to take medication, the process nodes for optional medicines, dosages, and side effects can be directly deleted; the focus can be placed solely on the correct diet list and the taboo diet list; in the correct diet list and the taboo diet list, the user can also adjust the order of foods in the above lists according to preference; after the process is determined, an executable automated process is generated and pushed to the user; the user can control his diet according to the process to achieve the purpose of adjusting his health and reducing the risk of disease.

[0040] S4: providing users with text, language, and / or visual charts for real-time interaction based on the generated automated process; The multimodal interaction module provides users with text, language and / or visual charts for real-time interaction based on the generated automated process; the multimodal interaction module interacts with users in real time, and users use voice and text to query and ask questions about the content displayed by the multimodal interaction module in real time; the latest output content is displayed; in the multimodal interaction module, users can choose to ask questions and have the system automatically process them and obtain the results of the questions; the form of questions asked by users can be text input or voice input; the results of the questions output by the system can also be in text, language; or in the form of charts to facilitate user understanding; in the multimodal interaction module, users can also ask questions again based on the output answers to the questions, and after obtaining satisfactory answers to the users, the RPA process management module is called for the user to choose to adjust the corresponding automated process.

[0041] S5: Optimize the knowledge graph and RPA process based on user behavior logs.

[0042] In this embodiment, the feedback optimization module is used to optimize the knowledge graph and RPA process based on user behavior logs. The feedback optimization module collects and analyzes user behavior logs during system usage, including the frequency of access to different content in health education resources, dwell time, interactive operation records, adjustment records of generated processes, and user feedback information; generates satisfaction scores, optimizes and updates the health knowledge graph based on the satisfaction level, supplements and improves the entities and relationships in the knowledge graph; makes targeted adjustments to the RPA process, and optimizes process steps and parameter settings. In this embodiment, a pop-up window can be used to invite users to provide a satisfaction rating before the output is completed, and this can be used as a reference. At the same time, corresponding scoring criteria are specified based on the degree of deletion and modification of the generated automated process by the user; for content that users actively delete, the corresponding weight is reduced when outputting the RPA process to reduce output. Example 3

[0043] This embodiment provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the knowledge graph-based health education resource RPA interaction method in Example 2 are implemented.

[0044] This embodiment also provides a computer-readable medium, which carries one or more programs. When the above one or more programs are executed by an electronic device, the electronic device implements the steps of the knowledge graph-based health education resource RPA interaction method in Example 2.

[0045] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned health education resource RPA interaction method based on knowledge graph have been referenced above. Figures 1 to 2The description of the knowledge graph-based health education resource RPA interaction system and method has been introduced in detail, and therefore, its repeated description will be omitted.

[0046] In summary, the RPA interaction method for health education resources based on the knowledge graph of the embodiment of the present application is explained, which integrates medical literature, electronic medical records and monitoring equipment data through a multi-source data fusion module, adopts data cleaning standardization to achieve entity alignment across data sources, and combines pre-trained models to complete entity recognition and relationship extraction, and constructs a health knowledge graph that supports dynamic node expansion; the graph-based RPA process management module generates a customizable automated health management process, and provides a visual adjustment interface to support users to add, delete, modify and sort process nodes; the multimodal interaction module integrates three interaction methods: text, voice and visual charts, to achieve two-way real-time interaction between natural language question and answer and process diagrams; the feedback optimization module collects user behavior logs and satisfaction data, and adopts machine learning methods to continuously optimize the entity relationships and RPA process parameters of the knowledge graph, forming a closed-loop system of "data-driven-intelligent generation-interaction optimization", and ultimately achieving accurate recommendation and dynamic evolution of personalized health education.

Claims

1. A health education resource RPA interactive system based on knowledge graph, characterized by: include: Multi-source data fusion module, RPA process management module, multimodal interaction module, and feedback optimization module; The multi-source data fusion module is used to aggregate health data from medical literature, electronic medical records, and monitoring equipment, and uses pre-trained models for entity recognition and relationship extraction to build a dynamically updated health knowledge graph; The RPA process management module is used to generate automated processes based on the health knowledge graph and allows users to dynamically adjust the generation rules; The multimodal interaction module provides users with text, language and / or visual charts for real-time interaction based on the generated automated process; The feedback optimization module is used to optimize the knowledge graph and RPA process based on user behavior logs.

2. The health education resource RPA interactive system based on knowledge graph according to claim 1 is characterized in that: The multi-source data fusion module cleans and standardizes the collected data, maps different terms for the same object in data from different sources to the same field, and unifies the units; the cleaned and standardized data is input into the pre-trained model for entity recognition and relationship extraction, thereby constructing the required health knowledge graph.

3. The health education resource RPA interactive system based on knowledge graph according to claim 1 is characterized in that: The RPA process management module generates an automated process based on the actual user input and the constructed health knowledge graph, and sets adjustment buttons in each node of the process, allowing users to adjust the order of process nodes, add or delete a node process, and modify the specific content of a node process.

4. The health education resource RPA interactive system based on knowledge graph according to claim 1 is characterized in that: The multimodal interaction module interacts with users in real time. Users can query and ask questions about the content displayed by the multimodal interaction module in real time through voice and text; and the latest output content is displayed.

5. The RPA interactive system for health education resources based on knowledge graph according to claim 1 is characterized in that: The feedback optimization module collects and analyzes user behavior logs during the use of the system, including the frequency of users' visits to different contents of health education resources, residence time, interactive operation records, adjustment records of the generated process, and user feedback information; generates satisfaction scores, optimizes and updates the health knowledge graph based on the satisfaction level, supplements and improves the entities and relationships in the knowledge graph; makes targeted adjustments to the RPA process, and optimizes process steps and parameter settings.

6. A knowledge graph-based RPA interaction method for health education resources, applied to the interactive system according to any one of claims 1 to 5, characterized in that: Including the following specific steps; S1: collect data and preprocess the data; S2: Based on the preprocessed data, the pre-trained model is called to perform entity recognition and relationship extraction to build a dynamically updated health knowledge graph; S3: Generate an automated process based on the user's actual input and the constructed health knowledge graph; S4: providing users with text, language, and / or visual charts for real-time interaction based on the generated automated process; S5: Optimize the knowledge graph and RPA process based on user behavior logs.

7. The RPA interaction method for health education resources based on knowledge graph according to claim 6 is characterized in that: In step S3, semantic recognition is performed based on the actual user input content to determine the user's required goals and potential goals; the goals are broken down into several subtasks that can be automated; and the subtasks are arranged in a logical order to output the final RPA process.

8. The RPA interaction method for health education resources based on knowledge graph according to claim 7 comprises: Analyze the subtasks broken down from the required goals and potential goals, and collaboratively process the subtasks with the same processing steps or types.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the knowledge graph-based health education resource RPA interaction method described in any one of claims 6 to 8 are implemented.

10. A computer-readable medium, characterized in that: The computer-readable medium carries one or more programs. When the one or more programs are executed by an electronic device, the electronic device implements the steps of the knowledge graph-based health education resource RPA interaction method described in any one of claims 6 to 8.

Images