An intelligent health consultation service method and system

Abstract

The application relates to an intelligent health consultation service method and system. The method comprises the following steps: acquiring multi-dimensional health consultation data, and adopting a mixed analysis strategy to perform information extraction on the health consultation data to obtain consultation information; based on the consultation information, combining a consultation mode rule, and matching a consultation channel; based on the consultation information, constructing or updating a health level pyramid model; based on the health level pyramid model, adopting the consultation channel, matching a health knowledge base, and generating a health service suggestion. The method can improve the accuracy and efficiency of health consultation.

Description

Technical Field

[0001] This invention belongs to the field of intelligent health consultation services, and in particular relates to an intelligent health consultation service method and system. Background Technology

[0002] With the rapid development of artificial intelligence and healthcare information technology, intelligent health consultation and home care technologies based on large-scale models are gradually becoming more widespread. These technologies support multimodal input, including text and images, and can achieve basic health Q&A and electronic record management. Traditional health care and consultation systems often employ manual user form filling for record creation, keyword matching and question-and-answer database responses, serial task processing, single-modal information parsing, and fixed-configuration resource scheduling. However, these traditional methods have many drawbacks: cumbersome record creation processes lead to low user compliance; rigid decision support lacks hierarchical and systematic health analysis; insufficient personalized services and inability to dynamically remember user information; weak multimodal processing capabilities, making it easy to create AI illusions when parsing medical images; low response efficiency; uneven resource allocation; lack of information verification mechanisms; difficulty in building accurate health profiles through natural interaction; and inability to meet users' needs for scientific, efficient, and personalized intelligent health care. Summary of the Invention

[0003] Therefore, it is necessary to provide an intelligent health consultation service method and system that can solve the above problems.

[0004] Firstly, this application provides a method for providing intelligent health consultation services, including:

[0005] Step 1: Obtain multi-dimensional health consultation data and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information;

[0006] Step 2: Based on the consultation information and in accordance with the consultation model rules, match the consultation channel;

[0007] Step 3: Based on the consultation information, construct or update the health hierarchy pyramid model;

[0008] Step 4: Based on the health hierarchy pyramid model, a consultation channel is used to match the health knowledge base and generate health service suggestions.

[0009] In one embodiment, the multi-dimensional health consultation data includes text input data and image upload data;

[0010] A hybrid analysis strategy was used to extract information from multi-dimensional health consultation data, resulting in consultation information, including:

[0011] For uploaded image data, an OCR engine is used to extract text data from the images;

[0012] The OCR engine includes a first processing channel, a second processing channel, and a fusion channel. The first processing channel is used to extract image recognition text based on uploaded image data. The second processing channel is used to identify the image type and the main object of the image based on uploaded image data, and generate the confidence scores for the image type and the main object. The fusion channel is used to verify the image recognition text, obtain the text verification result, and combine the text verification result, image type, main object, and the confidence scores for the image type and the main object to add labels to the image recognition text, forming image text data.

[0013] Semantic extraction is performed on image text data and text input data to obtain consultation semantic data;

[0014] Based on the consultation semantic mapping rules, the consultation semantic data is quantitatively mapped to obtain consultation semantic tags;

[0015] By integrating consultation semantic data and consultation semantic tags, consultation information is obtained.

[0016] In one embodiment, the health hierarchy pyramid model includes multiple target levels divided based on a preset health data dimension set, which includes basic health information dimension, symptom and sign dimension, disease risk dimension, diagnosis and treatment intervention dimension, and health management dimension.

[0017] Based on consultation information, construct or update the health hierarchy pyramid model, including:

[0018] Based on consultation information, extract basic health characteristic datasets;

[0019] Based on the preset health data dimensions, the basic health dataset is classified to generate a hierarchical matching feature set;

[0020] The hierarchical matching feature set is mapped to the corresponding target level in the health hierarchy pyramid model to obtain the basic hierarchical feature data.

[0021] Perform inter-level feature association verification on the basic hierarchical feature data and generate hierarchical association verification results;

[0022] Based on the basic hierarchical feature data, and according to the hierarchical association verification results, model update feature parameters are generated;

[0023] Based on the updated feature parameters of the model, a health hierarchy pyramid model can be constructed or an existing health hierarchy pyramid model can be updated.

[0024] In one embodiment, the consultation pattern rules include a consultation keyword filter, a first type of channel matching rules, a second type of channel matching rules, and a consultation channel set. Each consultation channel in the consultation channel set corresponds to a learning model. The learning model has unique reasoning ability and is pre-injected with specific response instructions.

[0025] Based on the consultation information and in accordance with the consultation model rules, a consultation channel is matched, including:

[0026] Consultation information is filtered using a consultation keyword filter to obtain a set of effective keywords, which is then used as the basis for matching the first type of channel.

[0027] Based on the first type of channel matching criteria, the first type of channel matching rules are used for matching to obtain the first type of channel matching results;

[0028] When the first type of channel matching result is no match, the consultation information is analyzed in context based on the effective keyword set to generate the second type of channel matching criteria. The second type of channel matching criteria includes consultation type, consultation urgency and consultation complexity.

[0029] Based on the second type of channel matching criteria, the second type of channel matching rules are used for matching to obtain the second type of channel matching results;

[0030] Based on the matching results of the first type of channel or the matching results of the second type of channel, the corresponding consultation channel is obtained from the consultation channel set.

[0031] In one embodiment, the method employs a parallel task execution architecture;

[0032] The parallel task execution system includes multiple pre-split independent tasks, predefined priority determination rules, pre-configured parallel execution rules based on timeout circuit breaking and degradation compensation, and a parallel task engine;

[0033] Multiple independent tasks are obtained by breaking down steps 1, 2, 3 and 4. Each independent task has predefined inter-task dependencies, required input data and filler input data.

[0034] The priority determination rules are configured to determine the execution priority of each independent task based on the dependencies between tasks and the required input data;

[0035] The parallel execution rules are configured to generate the parallel execution sequence of each independent task based on the execution priority;

[0036] When the parallel task execution system is executed, the corresponding independent tasks are called according to the parallel execution sequence. The initial execution result is obtained based on the required input data, and the initial execution result is incrementally completed based on the input data to obtain the final execution result.

[0037] In one embodiment, the formula for adding tags to image-recognized text is:

[0038]

[0039] in, Image text data, Recognize text from images. The confidence level of the text validation results. Confidence score for image type identification Confidence level for identifying the main object in an image. Assign confidence weights to the text test results. Assign confidence weights to image types. Assign confidence weights to the main subject of the image. For indicator functions, This is the confidence threshold for text validation.

[0040] In one embodiment, based on a health hierarchy pyramid model, a consultation channel is used to match a health knowledge base and generate health service recommendations, including:

[0041] Based on the hierarchical correlation verification results of each target level in the health hierarchy pyramid model, the core target level and related target levels are set.

[0042] Based on the core target hierarchy and related target hierarchy, and combined with the reasoning ability of the learning model corresponding to the consultation channel, hierarchical health knowledge data is obtained by matching from the health knowledge base, and hierarchical health service suggestions are generated by the learning model based on the hierarchical health knowledge data.

[0043] Perform a logical consistency check between hierarchical health service recommendations. Once the logical consistency check passes, generate health service recommendations.

[0044] Secondly, this application also provides an intelligent health consultation service system, including:

[0045] The data information extraction module is used to acquire multi-dimensional health consultation data and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information.

[0046] The consultation channel matching module is used to match consultation channels based on consultation information and consultation mode rules;

[0047] The health model building module is used to build or update a health hierarchy pyramid model based on consultation information.

[0048] The health advice generation module is used to generate health service suggestions based on the health hierarchy pyramid model, using a consultation channel and matching the health knowledge base.

[0049] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described intelligent health consultation service method.

[0050] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described intelligent health consultation service method.

[0051] The aforementioned intelligent health consultation service method and system, by acquiring multi-dimensional health consultation data and employing a hybrid analysis strategy to extract consultation information, can efficiently integrate text and image inputs and accurately analyze core health information, solving the problem of weak multimodal processing capabilities. Based on consultation information and consultation mode rules, it matches consultation channels and can dynamically allocate processing resources and response strategies according to problem characteristics, improving the problems of rigid resource allocation and low response efficiency. Based on consultation information, it constructs or updates a health hierarchy pyramid model, enabling hierarchical and systematic organization of health information and overcoming the shortcomings of rigid decision support. Based on this model and matching channels, it retrieves health knowledge bases to generate service suggestions, combining personalized health information to output targeted solutions, addressing pain points such as insufficient personalization and cumbersome record-keeping, and comprehensively improving the accuracy and efficiency of health consultations. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of an intelligent health consultation service method according to the present invention;

[0054] Figure 2 This is a structural diagram of an intelligent health consultation service system according to the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] In one embodiment, such as Figure 1 As shown, an intelligent health consultation service method is provided. This embodiment illustrates the application of this method to a consultation terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through the interaction between the terminal and the server. The real-time environment of this application includes: a terminal device such as a smartphone, tablet, or smart wearable device, used to collect and present the user's multimodal health consultation data; and a server, deployed with an analysis engine, rule base, and model library, used to execute data processing and decision-making logic. Application scenarios include: when a user has health consultation or personal health management needs, they input consultation information such as text and images through a terminal application. The terminal transmits the data to the server via the network. After receiving the data, the server performs hybrid analysis to extract consultation information, matches consultation channels according to rules, constructs or updates a health hierarchy pyramid model, and generates health service suggestions. The suggestions are then returned to the terminal interface for display through the matched channel, achieving efficient and personalized intelligent health care services.

[0057] In this embodiment, the method includes the following steps:

[0058] Step 1: Obtain multi-dimensional health consultation data and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information.

[0059] Optionally, multi-dimensional health consultation data refers to health-related information in different forms input by users through the interactive interface, which may include text input data (such as natural language text describing symptoms, medical history, and medication use in conversations) and image upload data (such as images taken or uploaded by users, such as physical examination reports, drug labels, and medical imaging reports). To achieve information extraction, the consultation terminal can adopt a hybrid analysis strategy, which can adapt and fuse different processing technology paths for different forms of data. For image upload data, the consultation terminal can extract text information through an optical character recognition engine; for example, this includes using an OCR model for high-precision text recognition to obtain image-recognized text, and calling image classification or object detection models to understand the image type (such as "lab report" or "prescription") and the subject object (such as "medicine box" or "wound"). The image understanding results can be used to assist in verification or provide context for subsequent analysis. The consultation terminal can perform quality checks on the text recognized by OCR (such as confidence-based filtering) and fuse it with the image understanding results to form image-text data. For text input data, the consultation terminal can employ natural language processing (NLP) techniques for semantic extraction. By combining rule-based fast matching (such as using regular expressions to extract structured fields like age and numerical values) with deep semantic understanding based on machine learning models (such as using named entity recognition and intent classification models to extract disease, symptom entities, and user consultation intent), health entities and user consultation intent can be parsed from the dialogue text, yielding consultation semantic data. The consultation terminal can integrate the text data of images with the semantic data of the input text to form unified, structured consultation information. This information includes key health entities, consultation intent, and related metadata (such as data source and confidence level) parsed from the user's input, providing input for subsequent decision analysis.

[0060] Step 2: Based on the consultation information and in accordance with the consultation model rules, match the consultation channel.

[0061] Optionally, the consultation terminal can input consultation information into a pre-configured intelligent scheduling and routing layer. This layer contains a pre-built set of consultation pattern rules, a collection of various logical judgment conditions and data structures used for qualitative analysis and classification of consultation requests. These rules include, but are not limited to: a keyword-based (e.g., "first aid," "pain," "record") fast filtering rule base for identifying high urgency or purely data-recording intents; and one or more lightweight machine learning models (e.g., classifiers) for more refined context-aware analysis of consultations that do not match the fast rules. This analysis can dynamically incorporate user health profile summaries and recent conversation history to comprehensively determine the deeper intent, complexity, and required professional depth of the consultation. Based on the analysis results of the consultation information according to the above consultation pattern rules (e.g., outputting classification tags such as "emergency help," "simple Q&A," and "in-depth health planning," along with complexity scores), the system executes a matching decision. The consultation channels point to multiple alternative processing service endpoints in the backend. Each channel is associated with an AI model instance or its configuration set that has specific reasoning capabilities and response styles (for example, the fast response channel corresponds to a standard model with low temperature parameters and short generation length, while the deep thinking channel corresponds to a dedicated model that supports chained reasoning and high generation length). The matching process involves selecting the most suitable channel from a predefined set based on the recommended patterns given by the analysis results, and routing the user request to the corresponding computing resources and processing logic to achieve adaptive alignment of response strategies, resource consumption, and problem difficulty.

[0062] Step 3: Based on the consultation information, construct or update the health hierarchy pyramid model.

[0063] Optionally, the health hierarchy pyramid model is a multi-layered data and knowledge logic framework for the structured representation and comprehensive analysis of an individual's health status. This health hierarchy pyramid model can consist of multiple logical levels divided according to preset health dimensions (e.g., a base layer representing the user's basic physiological attributes and medical history, a foundation layer reflecting their living environment and medical resources, a pillar layer corresponding to dynamic lifestyle interventions, a beam layer involving care behavior norms, and a roof layer pointing to long-term health management goals). The consultation terminal can dynamically integrate the consultation information extracted in step 1 (including entities such as symptoms, signs, examination values, and descriptions of lifestyle habits) into this model framework. The implementation process can include: the consultation terminal can use preset dimension-entity mapping rules to parse and classify the consultation information, identifying which target level(s) in the pyramid model the specific health elements involved should belong to. For example, when the consultation information includes "blood pressure 150 / 95," this data is classified and mapped to the body data dimension of the "base layer"; if it includes "a diet high in salt," it may be mapped to the dietary intervention dimension of the "pillar layer." After mapping is completed, the consultation terminal can perform model construction or update operations: if this is the first time a health profile has been created for the user, a health hierarchy pyramid model instance containing corresponding level data is initialized based on the mapping results; if the health hierarchy pyramid model already exists, the newly identified information is updated to the corresponding level's database fields in an incremental manner (such as adding, supplementing, or overwriting). This process may also include the verification and derivation of relationships between levels (for example, associating "advanced age" with "high blood pressure" to derive an enhancement in the "disease risk" dimension), transforming the health hierarchy pyramid model from a static container into a dynamic, live analysis framework that reflects the user's overall health profile and shows organic connections between dimensions, providing a structured basis for subsequent personalized decision-making.

[0064] Step 4: Based on the health hierarchy pyramid model, a consultation channel is used to match the health knowledge base and generate health service suggestions.

[0065] Optionally, the consultation terminal can determine one or more core target levels (such as the "symptoms and signs" level for acute symptom consultation) and logically related target levels (such as the "disease risk" level and "treatment intervention" level) based on the activation status and data completeness of each level in the constructed or updated health hierarchy pyramid model, combined with the immediate intent of the consultation. The consultation terminal can use the back-end processing unit corresponding to the consultation channel (such as an AI model configured with specific instruction templates and inference parameters) to perform knowledge retrieval and inference tasks. The consultation terminal can perform targeted retrieval and matching from the health knowledge base according to the determined target level; the health knowledge base is a structured aggregate of multi-source authoritative health knowledge (such as clinical guidelines, drug databases, nursing standards, nutritional advice, etc.), and its knowledge entries usually have a pre-defined correspondence with the pyramid model levels. The processing unit will integrate and infer the retrieved hierarchical health knowledge data with the personalized information (such as age, medical history, and recent physiological data) in the current user's health hierarchy pyramid model to generate preliminary "hierarchical health service suggestions" for each level. To ensure the integrity and scientific validity of the recommendations, the system can invoke preset verification rules to perform "inter-level logical consistency checks" on these hierarchical recommendations, eliminating inconsistencies and enhancing synergy (for example, ensuring that "exercise intervention" recommendations do not conflict with "underlying disease" restrictions). Finally, through integration, polishing, and formatting, a complete "health service recommendation" in natural language is generated and returned to the user, deeply integrating universal medical knowledge with the user's personalized health profile.

[0066] The aforementioned intelligent health consultation service method employs a hybrid analysis strategy in step 1 to extract information from text and image data, enabling the capture of structured health information within natural dialogue. This overcomes the pain points of "cumbersome profile creation and low user compliance" and "limited multimodal processing capabilities," laying the foundation for building accurate personal health profiles. Step 2, based on the extracted consultation information and rules, dynamically matches consultation channels, achieving intelligent identification and resource routing based on question complexity and intent. It adaptively aligns computational resources and response strategies with question difficulty, optimizing resource allocation, reducing response latency, and resolving issues such as "low response efficiency and rigid resource allocation." The problem of "systematization" is addressed in step 3. Step 3 utilizes the extracted information to dynamically construct or update a hierarchical health pyramid model, integrating fragmented health data into a scientific framework encompassing everything from basic physiology to long-term goals. This achieves a systematic and structured representation of the user's health status, fundamentally changing the current situation of "rigid decision support and lack of hierarchical analysis." Step 4 uses the aforementioned structured model as a blueprint to drive matching consultation channels that integrate personal profiles and authoritative knowledge bases to generate suggestions. This ensures that each output deeply integrates the user's personalized information and scientific evidence, improving health consultation efficiency in record-keeping, scientific decision-making, personalized services, and response speed.

[0067] In one embodiment, multi-dimensional health consultation data includes text input data and image upload data;

[0068] Optionally, multi-dimensional health consultation data includes text input data in the form of natural language text entered by the user through the interactive interface, as well as medical-related images taken or uploaded by the user, such as images of physical examination reports, prescriptions, or medicines.

[0069] A hybrid analysis strategy was used to extract information from multi-dimensional health consultation data to obtain consultation information, including:

[0070] S11 uses an OCR engine to extract text data from uploaded images.

[0071] The OCR engine includes a first processing channel, a second processing channel, and a fusion channel. The first processing channel is used to extract image recognition text based on uploaded image data. The second processing channel is used to identify the image type and the main object of the image based on uploaded image data, and generate the confidence scores for the image type and the main object. The fusion channel is used to verify the image recognition text, obtain the text verification results, and combine the text verification results, image type, main object, and the confidence scores for the image type and the main object to add labels to the image recognition text, forming image text data.

[0072] Optionally, for uploaded image data, the consultation terminal can invoke an Optical Character Recognition (OCR) engine optimized for medical documents to extract text information from the image. This OCR engine employs a dual-channel parallel processing architecture to improve efficiency and accuracy. The first processing channel, the text recognition channel, uses a deep learning-based OCR model to perform end-to-end text detection and recognition on the input image, outputting the original image-recognized text. The second processing channel, the image understanding channel, can be started in parallel. It invokes a pre-trained convolutional neural network model to perform image classification and object detection tasks on the same image. Its output includes a judgment of the overall image type (e.g., "blood test report," "CT report"), identification of key subjects in the image (e.g., "medicine box," "wound close-up"), and generation of corresponding image type confidence and subject object confidence. The results from both channels are fed into a subsequent fusion channel for comprehensive processing. The fusion channel can perform quality checks on the image-recognized text output by the first processing channel, for example, through vocabulary verification based on a medical professional dictionary and fluency analysis based on a statistical language model, generating a quantitative text verification result confidence score. The fusion channel executes the core fusion logic, weighting and synthesizing the text verification results, image type, image subject, and their respective confidence levels. Specifically, a pre-defined fusion formula adds a comprehensive quality and contextual identifier to the original image-to-text recognition. This formula is essentially a linearly weighted decision function, with weight coefficients configured based on prior knowledge in the medical field, such as assigning higher weights to the text's own verification results. The output is a structured image-text data object with a quantified confidence assessment. This identifier provides downstream processes with a clear measure of the reliability and relevance of the text data, realizing the transformation from raw image to reliable text data.

[0073] S12, semantic extraction is performed on the image text data and text input data to obtain consultation semantic data.

[0074] Optionally, the consultation terminal can perform unified semantic extraction on the obtained image text data and the original text input data. This process employs natural language processing technology, combining rule-based pattern matching with deep understanding based on machine learning models. The consultation terminal can perform named entity recognition, extracting key health entities from the text such as disease names, symptom descriptions, medications, dosages, test values, and physiological indicators. Simultaneously, it uses an intent classification model to determine the user's consultation intent, such as whether it is "information recording," "symptom consultation," or "treatment plan inquiry." This step generates structured consultation semantic data, containing a list of entities parsed from the user input, their relationships, and the core consultation intent.

[0075] S13. Based on the consultation semantic mapping rules, the consultation semantic data is quantitatively mapped to obtain consultation semantic labels.

[0076] Optionally, the consultation terminal can perform quantitative mapping of consultation semantic data based on predefined consultation semantic mapping rules. These mapping rules are typically a rule base or lightweight matching model that aligns extracted medical terms and symptom descriptions with standardized health concept classification systems or knowledge graph nodes. For example, the identified "hypertension" can be mapped to a disease entity label, and "150 / 95" can be mapped to a blood pressure value label and its corresponding severity level. Through this mapping, standardized, computable consultation semantic labels are added to the original semantic data.

[0077] S14, integrate consultation semantic data and consultation semantic tags to obtain consultation information.

[0078] Optionally, the consultation terminal can integrate consultation semantic data with metadata (such as source and confidence level) with structured consultation semantic tags to form a complete, standardized consultation information that can be directly used by subsequent decision-making modules, providing an accurate input basis for the construction of the health pyramid model and the matching of consultation channels.

[0079] In one embodiment, the health hierarchy pyramid model includes multiple target levels divided based on a preset health data dimension set, which includes basic health information dimension, symptom and sign dimension, disease risk dimension, diagnosis and treatment intervention dimension, and health management dimension.

[0080] Optionally, the health hierarchy pyramid model is a multi-layered data and knowledge logic framework for the structured representation and comprehensive analysis of individual health status. This model includes multiple target levels based on a pre-defined set of health data dimensions, which includes basic health information, symptoms and signs, disease risk, diagnosis and intervention, and health management. These five dimensions correspond to and construct the logical hierarchy in the pyramid model from basic facts to high-level goals.

[0081] Based on consultation information, construct or update the health hierarchy pyramid model, including:

[0082] S21. Based on consultation information, extract basic health feature dataset.

[0083] Optionally, the consultation terminal can extract a basic health feature dataset based on the extracted consultation information. The consultation information is structured data obtained after preliminary semantic extraction and mapping, containing health entities and their labels such as "70 years old," "hypertension," and "blood pressure 150 / 95." The consultation terminal can parse these structured entities and relationships, transforming them into a series of computable feature vectors or attribute key-value pairs to form the initial basic health feature dataset.

[0084] S22, classify the basic health dataset according to the preset health data dimensions, and generate a hierarchical matching feature set.

[0085] Optionally, the consultation terminal can classify the basic health feature dataset according to five preset health data dimensions. The consultation terminal has a pre-built dimension-entity mapping rule base or a lightweight classification model, which defines the affiliation relationships between various health concepts and the five dimensions. For example, features such as "age" and "gender" are classified into the "basic health information dimension," "dizziness" and "pain" into the "symptom and sign dimension," diagnostic conclusions such as "hypertension" and "diabetes" into the "disease risk dimension," suggestions such as "medication" and "rehabilitation training" into the "diagnosis and treatment intervention dimension," and intentions such as "long-term monitoring goals" and "lifestyle improvement" into the "health management dimension." Through this classification process, the original feature dataset is reorganized to generate a structured hierarchical matching feature set, where each feature is explicitly associated with its target level.

[0086] S23, map the hierarchical matching feature set to the corresponding target level in the health hierarchical pyramid model to obtain the basic hierarchical feature data.

[0087] Optionally, the consultation terminal can maintain a health hierarchy pyramid model instance for each user. This instance contains data storage areas corresponding to the five dimensions. The consultation terminal can write or associate each feature in the hierarchy matching feature set with the corresponding hierarchy node of the model instance, based on its categorized dimension. For example, the feature "blood pressure 150 / 95" (belonging to the symptom and sign dimension) can be added as a record to the "vital signs" subset under the "symptom and sign layer" of the model, thus obtaining updated basic hierarchy feature data containing the new features for this consultation.

[0088] S24, perform inter-level feature association verification on the basic hierarchical feature data, and generate hierarchical association verification results.

[0089] Optionally, the consultation terminal can perform inter-level feature association verification on the basic hierarchical feature data. This verification aims to ensure logical consistency between data at different levels and uncover potential associations. The consultation terminal can invoke a predefined medical logic rule engine or utilize knowledge graph relationships to check whether newly added features have known medical connections or contradictions with other existing hierarchical features in the model. For example, when "advanced age" (basic health information dimension) and "osteoporosis" (disease risk dimension) coexist, it can verify whether the weight of "fall prevention" (health management dimension) should be automatically increased or an association rule should be generated. This process will generate a hierarchical association verification result, which may contain newly discovered cross-level associations, conflict warnings, or derived implicit features.

[0090] S25, based on the basic hierarchical feature data, generates model update feature parameters according to the hierarchical association verification results.

[0091] Optionally, the consultation terminal can generate the final model update feature parameters based on the basic hierarchical feature data and combined with the hierarchical correlation verification results. These parameters determine whether the specific action of this update is to add, modify, supplement, or trigger derivative calculations. For example, if the verification results derive a new high-risk item, the update parameters will include an instruction to add this high-risk item as a derived feature to the "disease risk dimension".

[0092] S26. Based on the model update feature parameters, construct a health hierarchy pyramid model or update an existing health hierarchy pyramid model.

[0093] Optionally, the consultation terminal can update feature parameters based on the generated model to perform model construction or update operations. If the current user is consulting for the first time and a health hierarchy pyramid model has not yet been established, the consultation terminal initializes a brand new model instance according to the update parameters, and persistently stores all feature data extracted and verified in a hierarchical structure. If the model already exists, the system silently updates the corresponding level of the existing model with new features, correction values, or derived correlations in an incremental manner according to the update parameters, thereby completing a dynamic evolution of the user's health profile and ensuring that the pyramid model always reflects the latest and most systematic health status, providing a solid and structured basis for subsequent personalized decision-making.

[0094] In one embodiment, the consultation pattern rules include a consultation keyword filter, a first type of channel matching rules, a second type of channel matching rules, and a consultation channel set. Each consultation channel in the consultation channel set corresponds to a learning model. The learning model has unique reasoning ability and is pre-injected with specific response instructions.

[0095] Based on the consultation information and in accordance with the consultation model rules, a consultation channel is matched, including...

[0096] S31. Use a consultation keyword filter to filter consultation information to obtain a set of effective keywords, and use the set of effective keywords as the basis for matching the first type of channel.

[0097] Optionally, the consultation terminal can employ a consultation keyword filter. This filter is a predefined high-frequency keyword and pattern rule base, containing words with clear semantic meanings, such as "first aid," "pain," "record," and "hello," along with their combination rules. The consultation terminal can quickly match and scan the text content of the consultation information against this rule base, extracting all matching keywords to form a valid keyword set. This keyword set serves as the basis for the first type of channel matching.

[0098] S32, based on the first type of channel matching criteria, the first type of channel matching rules are used for matching to obtain the first type of channel matching result.

[0099] Optionally, the consultation terminal can invoke a first-type channel matching rule, which is a set of mapping rules based on deterministic logic. For example, it might define "if the keyword set contains both 'first aid' and a specific symptom term, then it will directly match to the emergency help channel." This rule is applied to the keyword set to determine the first-type channel matching result. If this result successfully matches a specific channel category (such as "emergency help," "simple greetings," or "data records"), the process terminates, and this result can be directly used as the routing basis.

[0100] S33. When the first type of channel matching result is no match, the consultation information is analyzed in context based on the effective keyword set to generate the second type of channel matching criteria. The second type of channel matching criteria includes consultation type, consultation urgency and consultation complexity.

[0101] Optionally, if the first type of matching result is "no match," it indicates that the user's question is relatively complex or the intent is implicit, and cannot be directly determined by simple keywords. In this case, the consultation terminal can initiate a deeper context-aware analysis. At this time, the consultation terminal can use the effective keyword set as clues, combined with the current consultation information (including the extracted health entities and preliminary intent) and dynamically injected context information (such as user health record summaries and recent dialogue history), and call a lightweight dedicated AI analysis model (i.e., an AI context understander) for parsing. The consultation terminal can integrate all information to make a multi-dimensional judgment on the consultation and output a structured second type of channel matching basis. Its core fields include consultation type (e.g., distinguishing between "fact-based question and answer," "diagnostic advice," or "deep planning"), consultation urgency (usually divided into high, medium, and low levels), and consultation complexity (assessed as simple, medium, or complex based on the breadth of the medical fields involved in the question and the length of the reasoning chain).

[0102] S34. Based on the second type of channel matching criteria, the second type of channel matching rules are used for matching to obtain the second type of channel matching result.

[0103] Optionally, the consultation terminal can use second-type channel matching rules to process the second-type channel matching criteria. The second-type channel matching rules are a set of policy-based decision logic or a lightweight classifier that maps the inputs of the three dimensions of type, urgency, and complexity to a specific "recommendation mode." For example, "high urgency + diagnostic advice" maps to "rapid professional consultation mode," and "high complexity + deep planning" maps to "deep thinking mode." This mapping result is the second-type channel matching result.

[0104] S35. Based on the matching results of the first type of channel or the matching results of the second type of channel, obtain the corresponding consultation channel from the consultation channel set.

[0105] Optionally, regardless of whether the matching result comes from the direct matching of the first type of rule or the inference matching result comes from the second type of rule, the consultation terminal can use both as input to obtain the corresponding consultation channel from the consultation channel set. The consultation channel set is a predefined resource configuration table, where each channel entry is associated with a specific backend learning model instance and its configuration. For example, the "fast response channel" is associated with a model that has a fast response speed and a limited number of generated tokens and has a pre-set instruction for concise answers, while the "deep planning channel" is associated with a model that supports complex inference, generates longer models, and has a pre-set instruction for structured output. The consultation terminal can select the corresponding channel configuration based on the matched pattern label, thereby completing accurate and adaptive routing from user questions to specific processing resources and response strategies.

[0106] In one embodiment, the method employs a parallel task execution architecture;

[0107] S41, the parallel task execution system includes multiple pre-split independent tasks, predefined priority determination rules, pre-configured parallel execution rules based on timeout circuit breaking and degradation compensation, and a parallel task engine;

[0108] S42, multiple independent tasks are obtained by breaking down steps 1, 2, 3 and 4. Each independent task has predefined inter-task dependencies, required input data and filler input data.

[0109] S43, The priority determination rule is configured to determine the execution priority of each independent task based on the dependencies between tasks and the required input data;

[0110] S44, the parallel execution rule is configured to generate the parallel execution sequence of each independent task based on the execution priority;

[0111] S45, when the parallel task execution system is executed, the corresponding independent task is called according to the parallel execution sequence, the initial execution result is obtained based on the required input data, and the initial execution result is incrementally completed based on the input data to obtain the final execution result.

[0112] Specifically, the parallel task execution system is a computing architecture designed to improve system processing efficiency and reliability. Its core is to reconstruct a complex, sequential processing flow (such as information extraction, model building, and channel matching steps in the preceding examples) into concurrently executable, fine-grained computing units. In implementation, the consultation terminal can analyze and deconstruct a complete business logic chain (e.g., the entire process from receiving user inquiries to generating a final response). For example, through modularization and dependency analysis, the process can be broken down into multiple logically independent computing modules represented by steps 1, 2, 3, and 4, each encapsulated as an independent task. Each independent task explicitly defines its inter-task dependencies (e.g., task B's startup must wait for task A's output), required input data (core data indispensable for task execution, such as task A's original output), and supplementary input data (auxiliary data used to enrich the task context and improve result quality, which can be provided asynchronously in subsequent stages, such as a summary of user history). The consultation terminal can schedule and sort these independent tasks according to predefined priority rules. This rule is a decision logic based on the topological sorting principle of directed acyclic graphs (DAG) and combined with the data preparation status. It is configured to prioritize independent tasks that have all the necessary input data ready and do not depend on other incomplete tasks (or whose dependent tasks have been completed), while taking into account the estimated computation time of the task itself, so as to ensure that the critical path tasks are executed first, and generate a preliminary task priority sequence.

[0113] Based on this priority sequence, parallel execution rules are activated to generate a precise parallel execution timing. This rule implements a fault-tolerance mechanism based on timeout circuit breaking and degradation compensation. The parallel execution rule allocates a computation window to each independent task and plans multiple truly parallel "task batches" according to their priority and dependencies. Simultaneously, the parallel execution rule presets a timeout threshold and degradation strategy for each task (e.g., when a data query task times out, it switches to using cached, possibly slightly older data, or triggers a simplified computation branch as compensation). The parallel task engine is the runtime environment that hosts this timing, typically consisting of a thread pool, task queue, and status monitor. When the parallel execution system starts executing, the engine can dynamically schedule eligible independent tasks to available computing resources for asynchronous execution based on the parallel execution timing. When each task is invoked, it performs core computations based on its currently obtained necessary input data, producing initial execution results. Afterward, the client does not block and wait but allows the task to continue execution. The consultation terminal asynchronously monitors and collects the arrival status of input data. Once the input data for a task is ready, a callback process for that task is immediately triggered to inject the input data and incrementally complete the initial execution results (for example, supplementing the user's past allergy history information to revise the recommendations based on preliminary diagnostic suggestions). This "first produce preliminary results based on necessary data, then iteratively enhance them using input data" model effectively reduces strong coupling and waiting delays between tasks. The entire execution process includes end-to-end state tracking. Task timeouts trigger preset circuit breakers and degradation compensation operations, ensuring that the overall process is not completely blocked due to anomalies in a single link. After all independent tasks are completed and data is completed, their final execution results are aggregated by the consultation terminal according to the business logic order, thus outputting the complete result of the entire process. This parallel execution system significantly improves processing throughput and response speed in complex business scenarios by parallelizing the process, atomicating tasks, explicitizing dependencies, and introducing elastic processing mechanisms, while ensuring service robustness.

[0114] In one embodiment, S51, the formula for adding a label to the image-recognized text is:

[0115]

[0116] in, Image text data, Recognize text from images. Confidence level of the text validation results. Confidence level for image type identification Confidence level for identifying the main object in an image. Assign confidence weights to the text test results. Assign confidence weights to image types. Assign confidence weights to the main subject of the image. For indicator functions, This is the confidence threshold for text validation.

[0117] For example, in this calculation formula, the original recognized text output by the optical character recognition (OCR) engine... By performing reliability assessment and weighted labeling, a final image and text data with a quality score can be generated that can be reliably used in subsequent processes. This calculation formula, by incorporating the confidence level and preset weights of multi-source evidence, achieves quantitative correction and quality control of the original identification results. The calculation formula... , and These represent the recognition confidence levels of three different dimensions: the text content itself, the overall image type, and the main object in the image. , and These are pre-defined weighting coefficients based on domain knowledge, used to adjust the influence of different dimensions of evidence on the final result. For example, in a healthcare scenario, the accuracy of the text content itself ( ) are usually given higher weights. The core calculation part of the formula This constitutes a weighted fusion function that linearly combines the confidence scores of the three independent sources, outputting a composite confidence score between 0 and 1. This score is compared with the original text. Multiplication, semantically equivalent to scaling the reliability of the original text, means that a higher score indicates a more refined and reliable text. The more reliable it is. Indicator function Acting as a safety valve in quality control, its logic is that the confidence level of the text inspection results is only considered if... Reaching the preset qualification threshold The entire expression is only valid (function value is 1) when the text quality is low; otherwise, the result is 0. This means that when the text itself is of low quality, regardless of other supporting evidence, the consultation terminal will discard the recognition result, fundamentally preventing low-quality or erroneous text from flowing into the downstream analysis process. Technically, this formula is a concrete digital manifestation of the "fusion channel" logic. During system operation, the OCR engine's "first processing channel" and "second processing channel" output in parallel. , and Then, the "fusion channel" calls the text verification module to check... Generate by performing validation (such as validation based on dictionaries, grammar, or medical terminology databases). Then, all these variables are substituted into this formula for calculation. Weighting coefficients and thresholds. It can be stored as a configurable parameter in the system configuration file. The calculated... Instead of unidentified raw strings, the data is now a structured data object with a quantified confidence assessment, which is then passed to the subsequent semantic extraction module. This design ensures that health information extracted from multimodal inputs, especially from images, has passed a quality filtering and confidence quantification check before entering the core health pyramid model construction and decision-making process. This significantly improves the data reliability of the consultation terminal when processing unstructured medical images, laying a data foundation for generating health recommendations.

[0118] In one embodiment, based on a health hierarchy pyramid model, a consultation channel is used to match a health knowledge base and generate health service recommendations, including:

[0119] S61, based on the hierarchical correlation verification results of each target level in the health hierarchy pyramid model, set the core target level and related target levels.

[0120] Optionally, the consultation terminal can analyze the correlation verification results between the target levels in the health hierarchy pyramid model, and based on this, set a core target level and several logically related target levels. The core target level points to the user's most prominent or urgent health needs (such as the management of a specific disease), while the related target levels cover other health dimensions closely related to the core need in terms of physiology, psychology, or behavior (such as nutritional support, exercise rehabilitation, or psychological adjustment related to the disease). In implementation, the consultation terminal can parse the hierarchical correlation network graph of the health hierarchy pyramid model, automatically determine the core level through a preset rule engine (e.g., based on the urgency of the need and weight scoring), and traverse the correlation edges to determine the correlation levels, thus defining a precise semantic scope for subsequent knowledge retrieval.

[0121] S62, based on the core target hierarchy and related target hierarchy, combined with the reasoning ability of the learning model corresponding to the consultation channel, obtains hierarchical health knowledge data from the health knowledge base, and generates hierarchical health service suggestions based on the hierarchical health knowledge data using the learning model.

[0122] Optionally, the consultation terminal can drive the consultation channel based on pre-defined core and related target hierarchies. Through this consultation channel, the corresponding learning model is invoked, utilizing the model's deep semantic understanding and reasoning capabilities to structurally query the health knowledge base. The query is not a simple keyword matching process; rather, the model transforms the hierarchical targets into a series of logically related query instructions and semantic constraints, matching and extracting hierarchically structured health knowledge data highly relevant to each level from the knowledge base. This data serves as the raw material for generating recommendations. Based on this hierarchical health knowledge data and its generation capabilities, the learning model transforms the knowledge points into specific, actionable, and naturally language-based preliminary recommendations, ensuring that the recommendations themselves maintain the hierarchical structure of core and related recommendations.

[0123] S63 performs a logical consistency check between hierarchical health service recommendations. After the logical consistency check passes, health service recommendations are generated.

[0124] Optionally, the consultation terminal can perform critical inter-level logical consistency checks on the generated hierarchical health service recommendations. This check is based on predefined inter-level constraint rules and common sense in the health knowledge base (for example, recommendations for "drug treatment" and "dietary nutrition" for a specific disease must not contain known conflicts of interest). The consultation terminal analyzes the recommendation text, identifies the health entities, behaviors, and quantitative indicators involved, and cross-compares the logical relationships between recommendations at different levels. Only when all identified logical relationships pass the rule base check is the set of recommendations considered logically consistent and integrated and formatted into a complete health service recommendation output. This ensures that the generated recommendations are both accurately targeted at multi-level health needs and constitute a coordinated and conflict-free overall solution from a medical logic perspective, meeting the reliability requirements of high-quality health services.

[0125] The aforementioned intelligent health consultation service method acquires multi-dimensional health consultation data and extracts consultation information using a hybrid analysis strategy. Specifically, for image data, a dual-channel OCR engine is used for extraction, and fusion computation is employed to add tags, effectively integrating text and image input, improving multimodal parsing accuracy, and suppressing AI illusions. This addresses the problems of cumbersome file creation, low user compliance, and weak multimodal processing capabilities inherent in traditional methods. Based on consultation information, and combined with consultation pattern rules including keyword filtering and contextual parsing, the method dynamically matches the consultation to a corresponding consultation channel of a learning model with specific reasoning capabilities. This achieves intelligent recognition of question complexity and intent, and adaptive resource routing, improving resource allocation efficiency. This approach addresses the shortcomings of inefficient decision support and response times. By constructing or updating a multi-dimensional health hierarchy pyramid model based on consultation information, including basic health information, symptoms, and signs, and verifying the correlation between levels, it achieves hierarchical and systematic organization and dynamic memorization of health information. This overcomes the pain points of rigid decision support, lack of systematic analysis, and insufficient personalized services. Finally, based on this pyramid model and matching consultation channels, hierarchical knowledge data is retrieved from the health knowledge base, suggestions are generated, and logical consistency between levels is verified. This ensures that the output health service suggestions deeply integrate personal health profiles and authoritative medical knowledge, thereby systematically improving the accuracy, personalization, and overall efficiency of health consultations.

[0126] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0127] Based on the same inventive concept, this application also provides an intelligent health consultation service system for implementing the intelligent health consultation service method described above. The solution provided by this system is similar to the implementation scheme described in the above method; therefore, the specific limitations of one or more embodiments of the intelligent health consultation service system provided below can be found in the limitations of the intelligent health consultation service method described above, and will not be repeated here.

[0128] In one exemplary embodiment, such as Figure 2 As shown, an intelligent health consultation service system is provided, including:

[0129] The data information extraction module 101 can be used to acquire multi-dimensional health consultation data and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information.

[0130] The consultation channel matching module 102 can be used to match consultation channels based on consultation information and consultation mode rules;

[0131] The health model building module 103 can be used to build or update a health hierarchy pyramid model based on consultation information.

[0132] The health advice generation module 104 can be used to generate health service suggestions based on the health hierarchy pyramid model, using a consultation channel and matching the health knowledge base.

[0133] In one embodiment, multi-dimensional health consultation data includes text input data and image upload data;

[0134] The data information extraction module 101 can also be used for:

[0135] For uploaded image data, an OCR engine is used to extract text data from the images;

[0136] The OCR engine includes a first processing channel, a second processing channel, and a fusion channel. The first processing channel is used to extract image recognition text based on uploaded image data. The second processing channel is used to identify the image type and the main object of the image based on uploaded image data, and generate the confidence scores for the image type and the main object. The fusion channel is used to verify the image recognition text, obtain the text verification result, and combine the text verification result, image type, main object, and the confidence scores for the image type and the main object to add labels to the image recognition text, forming image text data.

[0137] Semantic extraction is performed on image text data and text input data to obtain consultation semantic data;

[0138] Based on the semantic mapping rules for consultation, the semantic data of consultation is quantitatively mapped to obtain semantic labels for consultation.

[0139] By integrating consultation semantic data and consultation semantic tags, consultation information is obtained.

[0140] In one embodiment, the health hierarchy pyramid model includes multiple target levels divided based on a preset health data dimension set, which includes basic health information dimension, symptom and sign dimension, disease risk dimension, diagnosis and treatment intervention dimension, and health management dimension.

[0141] Health model building module 103 can also be used for:

[0142] Based on consultation information, extract basic health characteristic datasets;

[0143] Based on the preset health data dimensions, the basic health dataset is classified to generate a hierarchical matching feature set;

[0144] The hierarchical matching feature set is mapped to the corresponding target level in the health hierarchy pyramid model to obtain the basic hierarchical feature data.

[0145] Perform inter-level feature association verification on the basic hierarchical feature data and generate hierarchical association verification results;

[0146] Based on the basic hierarchical feature data, and according to the hierarchical association verification results, model update feature parameters are generated;

[0147] Based on the updated feature parameters of the model, a health hierarchy pyramid model can be constructed or an existing health hierarchy pyramid model can be updated.

[0148] In one embodiment, the consultation pattern rules include a consultation keyword filter, a first type of channel matching rules, a second type of channel matching rules, and a consultation channel set. Each consultation channel in the consultation channel set corresponds to a learning model. The learning model has unique reasoning ability and is pre-injected with specific response instructions.

[0149] The consultation channel matching module 102 can also be used for:

[0150] Consultation information is filtered using a consultation keyword filter to obtain a set of effective keywords, which is then used as the basis for matching the first type of channel.

[0151] Based on the first type of channel matching criteria, the first type of channel matching rules are used for matching to obtain the first type of channel matching results;

[0152] When the first type of channel matching result is no match, the consultation information is analyzed in context based on the effective keyword set to generate the second type of channel matching criteria. The second type of channel matching criteria includes consultation type, consultation urgency and consultation complexity.

[0153] Based on the second type of channel matching criteria, the second type of channel matching rules are used for matching to obtain the second type of channel matching results;

[0154] Based on the matching results of the first type of channel or the matching results of the second type of channel, the corresponding consultation channel is obtained from the consultation channel set.

[0155] In one embodiment, a parallel task execution module is included, for:

[0156] The method employs a parallel task execution architecture;

[0157] The parallel task execution system includes multiple pre-split independent tasks, predefined priority determination rules, pre-configured parallel execution rules based on timeout circuit breaking and degradation compensation, and a parallel task engine;

[0158] Multiple independent tasks are obtained by breaking down steps 1, 2, 3 and 4. Each independent task has predefined inter-task dependencies, required input data and filler input data.

[0159] The priority determination rules are configured to determine the execution priority of each independent task based on the dependencies between tasks and the required input data;

[0160] The parallel execution rules are configured to generate the parallel execution sequence of each independent task based on the execution priority;

[0161] When the parallel task execution system is executed, the corresponding independent tasks are called according to the parallel execution sequence. The initial execution result is obtained based on the required input data, and the initial execution result is incrementally completed based on the input data to obtain the final execution result.

[0162] In one embodiment, the health advice generation module 104 can also be used for:

[0163] Based on the hierarchical correlation verification results of each target level in the health hierarchy pyramid model, the core target level and related target levels are set.

[0164] Based on the core target hierarchy and related target hierarchy, and combined with the reasoning ability of the learning model corresponding to the consultation channel, hierarchical health knowledge data is obtained by matching from the health knowledge base, and hierarchical health service suggestions are generated by the learning model based on the hierarchical health knowledge data.

[0165] Perform a logical consistency check between hierarchical health service recommendations. Once the logical consistency check passes, generate health service recommendations.

[0166] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the intelligent health consultation service method as described above.

[0167] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0168] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0169] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for providing intelligent health consultation services, characterized in that, The method includes: Step 1: Obtain multi-dimensional health consultation data, and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information; Step 2: Based on the consultation information and in accordance with the consultation mode rules, match the consultation channel; Step 3: Based on the consultation information, construct or update the health hierarchy pyramid model; Step 4: Based on the health hierarchy pyramid model, use the consultation channel to match the health knowledge base and generate health service suggestions.

2. The method according to claim 1, characterized in that: The multi-dimensional health consultation data includes text input data and image upload data; The method employs a hybrid analysis strategy to extract information from the multi-dimensional health consultation data, obtaining consultation information, including: For the uploaded image data, an OCR engine is used to extract the text data from the image; The OCR engine includes a first processing channel, a second processing channel, and a fusion channel. The first processing channel is used to extract image recognition text based on the uploaded image data. The second processing channel is used to identify the image type and the main object of the image based on the uploaded image data, and generate the confidence scores for the image type and the main object of the image. The fusion channel is used to verify the image recognition text, obtain a text verification result, and combine the text verification result, the image type, the main object of the image, and the confidence scores for the image type and the main object of the image to add an identifier to the image recognition text, forming the image text data. Semantic extraction is performed on the image text data and the text input data to obtain consultation semantic data; Based on the consultation semantic mapping rules, the consultation semantic data is quantized and mapped to obtain consultation semantic tags; The consultation information is obtained by integrating the consultation semantic data and the consultation semantic tags.

3. The method according to claim 1, characterized in that: The health hierarchy pyramid model includes multiple target levels divided based on a preset health data dimension set, which includes basic health information dimension, symptom and sign dimension, disease risk dimension, diagnosis and treatment intervention dimension, and health management dimension. The construction or updating of the health hierarchy pyramid model based on the consultation information includes: Based on the consultation information, a basic health feature dataset is extracted; Based on the preset health data dimensions, the basic health dataset is classified to generate a hierarchical matching feature set; The hierarchical matching feature set is mapped to the corresponding target level in the health hierarchical pyramid model to obtain basic hierarchical feature data; Perform inter-level feature association verification on the basic hierarchical feature data to generate hierarchical association verification results; Based on the basic hierarchical feature data, and according to the hierarchical association verification results, model update feature parameters are generated; Based on the updated feature parameters of the model, the health level pyramid model is constructed or the existing health level pyramid model is updated.

4. The method according to claim 3, characterized in that: The consultation mode rules include a consultation keyword filter, a first type of channel matching rule, a second type of channel matching rule, and a consultation channel set. Each consultation channel in the consultation channel set corresponds to a learning model. The learning model has unique reasoning ability and is pre-injected with specific response instructions. The process of matching consultation channels based on the consultation information and consultation mode rules includes: The consultation information is filtered using a consultation keyword filter to obtain a set of effective keywords, and this set of effective keywords is used as the basis for the first type of channel matching. Based on the first type of channel matching criteria, the first type of channel matching rules are used for matching to obtain the first type of channel matching result; When the first type of channel matching result is no match, the consultation information is analyzed in context based on the effective keyword set to generate a second type of channel matching basis. The second type of channel matching basis includes consultation type, consultation urgency and consultation complexity. Based on the second type of channel matching criteria, the second type of channel matching rules are used for matching to obtain the second type of channel matching results; Based on the matching results of the first type of channel or the matching results of the second type of channel, the corresponding consultation channel is obtained from the consultation channel set.

5. The method according to claim 1, characterized in that: The method employs a parallel task execution architecture; The parallel task execution system includes multiple pre-split independent tasks, predefined priority determination rules, pre-configured parallel execution rules based on timeout circuit breaking and degradation compensation, and a parallel task engine; The multiple independent tasks are obtained by breaking down steps 1, 2, 3 and 4. Each independent task has a predefined inter-task dependency relationship, required input data and filler input data. The priority determination rule is configured to determine the execution priority of each independent task based on the inter-task dependencies and the required input data; The parallel execution rule is configured to generate the parallel execution sequence of each of the independent tasks based on the execution priority; When the parallel task execution system is executed, the corresponding independent task is called according to the parallel execution sequence, the initial execution result is obtained based on the required input data, and the initial execution result is incrementally completed based on the filling input data to obtain the final execution result.

6. The method according to claim 2, characterized in that, The formula for adding identifiers to the image-recognized text is: in, The image text data, Recognize text in the image. Confidence level of the text validation results. Confidence score for image type identification Confidence level for identifying the main object in an image. Assign confidence weights to the text test results. Assign confidence weights to image types. Assign confidence weights to the main subject of the image. For indicator functions, This is the confidence threshold for text validation.

7. The method according to claim 4, characterized in that, Based on the aforementioned health hierarchy pyramid model, the consultation channel is used to match the health knowledge base and generate health service suggestions, including: Based on the hierarchical correlation verification results of each target level in the health hierarchy pyramid model, the core target level and the associated target level are set. Based on the core target hierarchy and the associated target hierarchy, and combined with the reasoning ability of the learning model corresponding to the consultation channel, hierarchical health knowledge data is obtained from the health knowledge base, and hierarchical health service suggestions are generated using the learning model based on the hierarchical health knowledge data. The hierarchical health service recommendations are subjected to inter-hierarchical logical consistency verification. After the logical consistency verification passes, the health service recommendations are generated.

8. An intelligent health consultation service system, characterized in that, The system includes: The data information extraction module is used to acquire multi-dimensional health consultation data and use a hybrid analysis strategy to extract information from the health consultation data to obtain consultation information. The consultation channel matching module is used to match consultation channels based on the consultation information and consultation mode rules. A health model building module is used to build or update a health hierarchy pyramid model based on the consultation information. The health advice generation module is used to generate health service advice based on the health hierarchy pyramid model, using the consultation channel, and matching the health knowledge base.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images