A remote medical health consultation and management platform

Abstract

The present application relates to the field of intelligent consultation, and specifically discloses a remote medical health consultation and management platform, which includes: a user detection terminal, a user terminal, a health management platform background server, and an intelligent question-and-answer terminal. First, the user detection terminal collects body index data, and the user terminal receives the text description of the consultation question. This information is stored and managed by the health management platform background server, and then the intelligent question-and-answer terminal uses the collected data to provide users with customized answers based on their personal health status and specific questions, thereby realizing the instant generation of consultation answer texts and providing more personalized consultation services.

Description

Technical Field

[0001] The present application relates to the field of intelligent consultation, and more specifically, to a remote medical health consultation and management platform. Background Art

[0002] With the rapid development of Internet technology and the popularity of smart phones, telemedicine services have become an increasingly popular healthcare model. Telemedicine services can break through geographical restrictions and allow patients to obtain professional medical consultation and health guidance at home, which is particularly important in remote areas or emergency situations.

[0003] Chinese patent CN106096280A proposes a remote medical health consultation and management platform, which includes a user detection terminal, a user terminal, a backend server, a doctor terminal, a health consultant terminal, and a drug and device retail terminal. Specifically, the user sends physiological data to the server through the network, and the doctor evaluates and feeds back to the platform to generate a health report. The platform provides health advice to the user based on the doctor's opinion and supports video calls between the user and the health consultant.

[0004] The above patents mainly rely on doctor terminals to provide consulting opinions to users. However, doctor resources are relatively limited, especially in small cities or remote areas, which themselves face a shortage of medical talents. When relying on doctor terminals to provide services, if doctors need to review the health data of each user one by one and make a detailed assessment, this will greatly increase the doctor's workload, resulting in longer service response time, and the service quality may also decline. In addition, doctors need to review the data and reply within the specified time, which means that while dealing with daily work tasks, they also have to deal with online consultations, which can easily cause fatigue and affect judgment. In an emergency, this delay may lead to the missing of critical treatment opportunities, especially for those health problems that need to be dealt with immediately, thereby increasing health risks.

[0005] Therefore, an optimized remote medical health consultation and management platform is desired. Summary of the invention

[0006] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides a remote medical health consultation and management platform, which obtains the user's body indicator data collected by the user detection terminal, and obtains the consultation question text description entered by the user, and adopts natural language processing technology and large language models. First, find the template question that matches the consultation question text description, then semantically encode the consultation question text description and the template question, and then conditionally guide the interaction between the original semantic features of the consultation question and the semantic features of the template question to obtain the enhanced semantic features of the consultation question, and at the same time, globally embed the user's body indicator data, so as to intelligently obtain the consultation answer text according to the fusion semantics between the global embedded semantic features of the body indicators. In this way, the consultation answer text can be generated instantly, reducing the user's waiting time, and at the same time, it can provide more personalized health advice, assist doctors in medical and health consultation work, reduce the workload of doctors, and provide more personalized consulting services.

[0007] According to one aspect of the present application, a remote medical health consultation and management platform is provided, which includes:

[0008] The user detection terminal is used to collect the user's physical index data; the user terminal is used to receive the consultation question text description input by the user; the health management platform backend server is used to store user information; the remote medical health consultation and management platform also includes: an intelligent question and answer terminal, which is used to obtain the consultation answer text based on the consultation question text description input by the user and the user's physical index data; wherein the intelligent question and answer terminal includes:

[0009] A user information acquisition module, used to acquire the user's physical index data collected by the user detection terminal, and acquire the text description of the consultation question input by the user;

[0010] A text description semantic encoding module, used for obtaining a template question matching the text description of the consulting question based on the text description of the consulting question, and performing semantic encoding on the text description of the consulting question and the template question matching the text description of the consulting question to obtain original semantic features of the consulting question and semantic features of the template question;

[0011] A consulting question enhancement module, used for performing conditionally guided inter-feature significant interaction on the original semantic features of the consulting question and the semantic features of the template question to obtain an enhanced semantic representation of the consulting question;

[0012] A question body index splicing enhancement module is used to embed and encode the user's body index data, and then splice the obtained body index global embedding semantic features and the consultation question enhanced semantic representation to obtain a consultation question-body index splicing semantic representation;

[0013] The consultation question text generation module is used to obtain the consultation answer text based on the consultation question-physical indicator splicing semantic representation.

[0014] In the above-mentioned telemedicine health consultation and management platform, the text description semantic encoding module is used to: based on the text description of the consulting question, extract the template question that matches the text description of the consulting question from the template question database; semantically encode the text description of the consulting question and the template question that matches the text description of the consulting question to obtain the original semantic encoding vector of the consulting question as the original semantic feature of the consulting question and the template question semantic encoding vector as the template question semantic feature.

[0015] In the above-mentioned telemedicine health consultation and management platform, the consulting question enhancement module includes: a consulting question-template question latent variable feature combination subunit, which is used to perform latent variable feature combination on the original semantic coding vector of the consulting question and the semantic coding vector of the template question to obtain a consulting question-template question conditional feature vector; a consulting question-template question feature interaction subunit, which is used to perform conditionally dependent feature interaction fusion on the original semantic coding vector of the consulting question and the semantic coding vector of the template question based on the consulting question-template question conditional feature vector to obtain the enhanced semantic representation of the consulting question.

[0016] In the above-mentioned telemedicine health consultation and management platform, the consulting question-template question latent variable feature joint subunit includes: a consulting question-template question semantic joint secondary subunit, which is used to input the consulting question original semantic coding vector and the template question semantic coding vector into the joint implicit feature capture network to obtain a consulting question-template question semantic joint implicit feature vector; a consulting question-template question conditional feature generation secondary subunit, which is used to perform feature activation based on the Sigmoid function on the consulting question-template question semantic joint implicit feature vector to obtain the consulting question-template question conditional feature vector.

[0017] In the above-mentioned telemedicine health consultation and management platform, the consulting question-template question semantic joint secondary sub-unit is used to: after adding the original semantic coding vector of the consulting question and the semantic coding vector of the template question by position, multiply the obtained consulting question-template question semantic sum vector with the weight matrix and then add it by position with the bias vector to obtain the consulting question-template question semantic joint interaction vector; use the tanh function to process the consulting question-template question semantic joint interaction vector to obtain the consulting question-template question semantic joint implicit feature vector.

[0018] In the above-mentioned telemedicine health consultation and management platform, the consulting question-template question feature interaction subunit includes: a consulting question semantic contribution calculation secondary subunit, which is used to calculate the consulting question semantic contribution of the consulting question original semantic coding vector relative to the consulting question-template question conditional feature vector; a template question semantic contribution calculation secondary subunit, which is used to calculate the template question semantic contribution of the template question semantic coding vector relative to the consulting question-template question conditional feature vector; a consulting template question semantic modulation secondary subunit, which is used to normalize the consulting question semantic contribution and the template question semantic contribution, and use the normalized consulting question semantic contribution and the normalized template question semantic contribution to weighted modulate the consulting question original semantic coding vector and the template question semantic coding vector to obtain the modulated consulting question original semantic coding vector and the modulated template question semantic coding vector; a consulting question enhancement secondary subunit, which is used to use the modulated consulting question original semantic coding vector as the query vector and the modulated template question semantic coding vector as the query vector and the modulated template question semantic coding vector as the query vector. as the key vector and the consulting question-template question conditional feature vector as the value vector, the modulated consulting question original semantic coding vector, the modulated template question semantic coding vector and the consulting question-template question conditional feature vector are input into the feature-to-feature significant guided interaction module based on the converter structure to obtain the consulting question enhanced semantic coding vector as the consulting question enhanced semantic representation; wherein the consulting question semantic contribution degree calculation secondary subunit is used to: calculate the positional division of the consulting question original semantic coding vector and the consulting question-template question conditional feature vector to obtain the consulting question semantic contribution vector; calculate the logarithmic function value of the absolute value of each eigenvalue of the consulting question semantic contribution vector with base two to obtain the consulting question semantic contribution logarithmic vector; calculate the positional dot product of the consulting question original semantic coding vector and the consulting question semantic contribution logarithmic vector, and perform position-by-position multiplication on the obtained dot product vector to obtain the consulting question semantic contribution value; calculate the exponential function with the natural constant e as the base and the consulting question semantic contribution value as the exponent to obtain the consulting question semantic contribution degree.

[0019] In the above-mentioned telemedicine health consultation and management platform, the consultation template question semantic modulation secondary sub-unit is used to: calculate the sum of the consultation question semantic contribution and the template question semantic contribution to obtain the consultation question-template question semantic contribution sum; respectively divide the consultation question semantic contribution and the template question semantic contribution by the consultation question-template question semantic contribution sum to obtain the normalized consultation question semantic contribution and the normalized template question semantic contribution; perform positional point multiplication on the consultation question original semantic coding vector and the normalized consultation question semantic contribution to obtain the modulated consultation question original semantic coding vector; perform positional point multiplication on the template question semantic coding vector and the normalized template question semantic contribution to obtain the modulated template question semantic coding vector.

[0020] In the above-mentioned remote medical health consultation and management platform, the consulting question enhancement secondary sub-unit is used to: perform vector multiplication on the modulated original semantic coding vector of the consulting question and the transposed vector of the modulated template question semantic coding vector, and divide the obtained modulated consulting question-template question semantic association matrix by the square root of the length of the modulated template question semantic coding vector by position to obtain the modulated consulting question-template question semantic association scaling matrix; input the modulated consulting question-template question semantic association scaling matrix into the softmax function, and perform matrix-vector multiplication on the obtained modulated consulting question-template question semantic association scaling activation matrix and the consulting question-template question conditional feature vector to obtain the consulting question enhanced semantic coding vector.

[0021] In the above-mentioned remote medical health consultation and management platform, the question body indicator splicing enhancement module is used to: use the body indicator embedding coding matrix to embed the user's various body indicator data to obtain a body indicator global embedded semantic coding vector as the body indicator global embedded semantic feature; splice the consulting question enhanced semantic coding vector and the body indicator global embedded semantic coding vector to obtain a consulting question-body indicator splicing semantic coding vector as the consulting question-body indicator splicing semantic representation.

[0022] In the above-mentioned remote medical health consultation and management platform, the consultation question text generation module is used to: input the consultation question-physical index splicing semantic coding vector into the intelligent consultation module based on the large language model to obtain the consultation answer text.

[0023] Compared with the prior art, the present application provides a remote medical health consultation and management platform, which obtains the user's body indicator data collected by the user detection terminal, and obtains the consultation question text description entered by the user, and adopts natural language processing technology and large language models. First, find the template question that matches the consultation question text description, then semantically encode the consultation question text description and the template question, and then conditionally guide the interaction between the original semantic features of the consultation question and the semantic features of the template question to obtain the enhanced semantic features of the consultation question, and at the same time, globally embed the user's body indicator data, so as to intelligently obtain the consultation answer text according to the fusion semantics between the global embedded semantic features of the body indicators. In this way, the consultation answer text can be generated instantly, reducing the user's waiting time, and at the same time, it can provide more personalized health advice, assist doctors in medical and health consultation work, reduce the workload of doctors, and provide more personalized consulting services. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0025] Figure 1 4 is a block diagram of a remote medical health consultation and management platform according to an embodiment of the present application.

[0026] Figure 2 It is a block diagram of an intelligent question-and-answer terminal in a remote medical health consultation and management platform according to an embodiment of the present application.

[0027] Figure 3 Schematic diagram of data flow of the intelligent question-and-answer terminal in the remote medical health consultation and management platform according to an embodiment of the present application.

[0028] Figure 4 This is a block diagram of a consultation question enhancement module in a remote medical health consultation and management platform according to an embodiment of the present application.

[0029] Figure 5 It is a block diagram of the consultation question-template question latent variable feature joint subunit in the remote medical health consultation and management platform according to the embodiment of the present application. DETAILED DESCRIPTION

[0030] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0031] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0032] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0033] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0034] It should also be noted that in this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0035] With the rapid development of Internet technology and the popularity of smart phones, telemedicine services have become an increasingly popular healthcare model. Telemedicine services can break through geographical restrictions and allow patients to obtain professional medical consultation and health guidance at home, which is particularly important in remote areas or emergency situations.

[0036] Chinese patent CN106096280A proposes a remote medical health consultation and management platform, which includes a user detection terminal, a user terminal, a backend server, a doctor terminal, a health consultant terminal, and a drug and equipment retail terminal. Specifically, the user obtains physiological data through the network to the server, and the doctor evaluates and feeds back to the platform to generate a health report. The platform provides health advice to the user based on the doctor's opinion and supports video calls between the user and the health consultant.

[0037] The above patents mainly rely on doctor terminals to provide consulting opinions to users. However, doctor resources are relatively limited, especially in small cities or remote areas, which themselves face a shortage of medical talents. When relying on doctor terminals to provide services, if doctors need to review the health data of each user one by one and make a detailed assessment, this will greatly increase the doctor's workload, resulting in longer service response time, and the service quality may also decline. In addition, doctors need to review the data and reply within the specified time, which means that while dealing with daily work tasks, they also have to deal with online consultations, which can easily cause fatigue and affect judgment. In an emergency, this delay may lead to the missing of critical treatment opportunities, especially for those health problems that need to be dealt with immediately, thereby increasing health risks.

[0038] Based on this, the present application proposes a remote medical health consultation and management platform, which introduces an intelligent question-and-answer terminal to achieve instant response and auxiliary guidance, thereby reducing the workload of doctors and providing more convenient health consultation services. Figure 1 FIG. 1 is a block diagram of a remote medical health consultation and management platform according to an embodiment of the present application. Figure 1 As shown, the remote medical health consultation and management platform 100 according to the embodiment of the present application specifically includes: a user detection terminal 110, used to collect various physical indicator data of the user; a user terminal 120, used to receive a text description of a consultation question input by the user; a health management platform backend server 130, used to store user information; and an intelligent question and answer terminal 140, which is used to obtain a consultation answer text based on the consultation question text description input by the user and the various physical indicator data of the user.

[0039] Correspondingly, in the intelligent question-and-answer terminal, the technical concept of the present application is to obtain the user's body indicator data collected by the user detection terminal, and obtain the consultation question text description entered by the user, and adopt natural language processing technology and large language models, first find the template question that matches the consultation question text description, then semantically encode the consultation question text description and the template question, and then conditionally guide the interaction between the original semantic features of the consultation question and the semantic features of the template question to obtain the enhanced semantic features of the consultation question, and at the same time globally embed the user's body indicator data, so as to intelligently obtain the consultation answer text according to the fusion semantics between the global embedded semantic features of the body indicators. In this way, the consultation answer text can be generated instantly, reducing the user's waiting time, and at the same time, it can provide more personalized health advice, assist doctors in medical and health consultation work, reduce the workload of doctors, and provide more personalized consulting services.

[0040] Figure 2 It is a block diagram of an intelligent question-and-answer terminal in a remote medical health consultation and management platform according to an embodiment of the present application. Figure 3 Schematic diagram of data flow of intelligent question-answering terminal in remote medical health consultation and management platform according to the embodiment of the present application. Figure 2 and Figure 3 As shown, the intelligent question-and-answer terminal 140 includes: a user information acquisition module 141, which is used to obtain the user's physical indicator data collected by the user detection terminal, and obtain the text description of the consulting question input by the user; a text description semantic encoding module 142, which is used to obtain a template question matching the consulting question text description based on the consulting question text description, and semantically encode the consulting question text description and the template question matching the consulting question text description to obtain the original semantic features of the consulting question and the semantic features of the template question; a consulting question enhancement module 143, which is used to perform conditionally guided inter-feature significant interactions on the original semantic features of the consulting question and the semantic features of the template question to obtain an enhanced semantic representation of the consulting question; a question body indicator splicing enhancement module 144, which is used to embed and encode the user's physical indicator data, and then splice the obtained global embedded semantic features of the body indicators with the enhanced semantic representation of the consulting question to obtain a consulting question-body indicator splicing semantic representation; a consulting question text generation module 145, which is used to obtain the consulting answer text based on the consulting question-body indicator splicing semantic representation.

[0041] In the embodiment of the present application, the user information acquisition module 141 is used to obtain the user's body index data collected by the user detection terminal, and obtain the consultation question text description entered by the user. It should be understood that the body index data are the user's physiological parameters collected by various wearable devices or home medical devices (such as smart watches, sphygmomanometers, blood glucose meters, etc.), such as heart rate, blood pressure, blood oxygen saturation, body temperature, blood sugar level, etc., which are used to provide the user's basic health status. The consultation question text description entered by the user refers to the text information entered by the user through the interface provided by the platform, which contains the specific questions or symptom descriptions that the user wants to consult. For example, the user may describe that he feels tired, has headache symptoms, or asks about the causes and coping methods of a specific symptom. Based on this, in order to be able to generate personalized consultation texts based on the user's situation more accurately, in the technical solution of the present application, the user's body index data collected by the user detection terminal are obtained, and the consultation question text description entered by the user is obtained.

[0042] Specifically, in the embodiment of the present application, the text description semantic encoding module 142 is used to obtain a template question matching the text description of the consulting question based on the text description of the consulting question, and to semantically encode the text description of the consulting question and the template question matching the text description of the consulting question to obtain the original semantic features of the consulting question and the semantic features of the template question. Specifically, in the embodiment of the present application, the text description semantic encoding module is used to: extract the template question matching the text description of the consulting question from the template question database based on the text description of the consulting question; semantically encode the text description of the consulting question and the template question matching the text description of the consulting question to obtain the original semantic encoding vector of the consulting question as the original semantic features of the consulting question and the semantic encoding vector of the template question as the semantic features of the template question. Accordingly, considering that the text description of the consulting question contains the user's consulting intention and the category information of the question, and in order to quickly identify the category and intention of the user's question, thereby speeding up the problem processing, while ensuring the professionalism and accuracy of the answer, in the technical solution of the present application, based on the text description of the consulting question, extract the template question matching the text description of the consulting question from the template question database. In particular, the template questions in the template question database are usually carefully designed and verified, and can cover most common problems, including specific medical terms and clinical scenarios. By proposing corresponding template questions from the template question database, the consistency and standardization of answers can be ensured, the inconsistent answers caused by the uncertainty of natural language processing can be avoided, and misunderstandings and errors can be reduced. Then, considering that the consulting question text description and the template questions matching the consulting question text description contain rich semantic information, such as semantic associations between contexts, emotional tendencies, etc. Based on this, in order to better capture the contextual semantic relationship between words or phrases, in the technical solution of the present application, the consulting question text description and the template questions matching the consulting question text description are semantically encoded to obtain the original semantic encoding vector of the consulting question and the template question semantic encoding vector, so that these complex semantic information can be understood and processed more clearly, providing data support for subsequent data interaction. In particular, in one embodiment of the present application, the consulting question text description and the template questions matching the consulting question text description can be semantically encoded by a semantic encoder comprising an embedding layer to capture and mine the associated semantic information.

[0043] In the embodiment of the present application, the consulting question enhancement module 143 is used to perform conditionally guided significant interaction between the original semantic features of the consulting question and the semantic features of the template question to obtain an enhanced semantic representation of the consulting question. Figure 4: is a block diagram of a consultation question enhancement module in a remote medical health consultation and management platform according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 4 As shown, the consulting question enhancement module 143 includes: a consulting question-template question latent variable feature combination subunit 1431, which is used to perform latent variable feature combination on the original semantic coding vector of the consulting question and the semantic coding vector of the template question to obtain a consulting question-template question conditional feature vector; a consulting question-template question feature interaction subunit 1432, which is used to perform conditional dependent feature interaction fusion on the original semantic coding vector of the consulting question and the semantic coding vector of the template question based on the consulting question-template question conditional feature vector to obtain the enhanced semantic representation of the consulting question. Accordingly, considering that the original semantic coding vector of the consulting question contains the contextual semantic information in the user's question, it can help understand the user's actual needs and specific situations. The template question semantic coding vector reflects the basic structure and format of the template question, usually covers the core elements of a certain type of question, and provides a standardized reference framework. At the same time, there is a mutual implicit association relationship between the two in semantics. Therefore, in order to further improve the accuracy of understanding and the quality of answering users' consulting questions, in the technical solution of the present application, a significant interaction between the features of the original semantic coding vector of the consulting question and the semantic coding vector of the template question is conditionally guided to obtain an enhanced semantic coding vector of the consulting question as an enhanced semantic representation of the consulting question.

[0044] Specifically, first, the original semantic coding vector of the consulting question and the semantic coding vector of the template question are combined with latent variable features to better reflect the implicit relationship between the user question and the standard template question, and the consulting question-template question conditional feature vector is obtained. Specifically, the original semantic coding vector of the consulting question and the semantic coding vector of the template question are jointly captured using a joint implicit feature capture network to learn the intrinsic semantic connection between the two sets of input features, while revealing the implicit information in the original features, and obtaining a consulting question-template question semantic joint implicit feature vector. Afterwards, in order to further enhance and highlight the most critical part of the comprehensive conditional features, the Sigmoid function is used to activate the joint implicit feature vector to obtain the consulting question-template question conditional feature vector.

[0045] Next, based on the consulting question-template question conditional feature vector, the original semantic coding vector of the consulting question and the semantic coding vector of the template question are subjected to conditionally dependent feature interactive fusion to obtain an enhanced semantic coding vector for the consulting question. This can better capture the deep semantic information of the user's consulting question and combine it with the standard semantic information in the template question, thereby enhancing the understanding of the core of the question. Specifically, the semantic contribution of the original semantic coding vector of the consulting question and the semantic coding vector of the template question relative to the consulting question-template question conditional feature vector are calculated respectively to evaluate the correlation of each feature relative to the conditional vector in order to determine the importance and contribution of each feature to the final representation, and obtain the semantic contribution of the consulting question and the semantic contribution of the template question. Then, in order to make all weights on the same scale to ensure the fairness of subsequent weighted modulation, the semantic contribution of the consulting question and the semantic contribution of the template question are normalized, and the normalized semantic contribution of the consulting question and the normalized semantic contribution of the template question are weighted modulated on the original semantic encoding vector of the consulting question and the semantic encoding vector of the template question, so as to effectively emphasize those parts that are more important and critical to the conditional feature vector, and appropriately reduce the influence and interference of irrelevant factors. Finally, the original semantic encoding vector of the consulting question after modulation is used as the query vector, the semantic encoding vector of the template question after modulation is used as the key vector, and the conditional feature vector of the consulting question-template question is used as the value vector. The significant interaction module between the features based on the converter structure is used to achieve significant interaction between different features, so that the model can better understand the contextual semantic relationship between features and focus on the most important and critical features, and finally obtain the enhanced semantic encoding vector of the consulting question.

[0046] Figure 5 : is a block diagram of a consulting question-template question latent variable feature joint subunit in a remote medical health consultation and management platform according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 5 As shown, the consulting question-template question latent variable feature joint subunit 1431 includes: a consulting question-template question semantic joint secondary subunit 14311, which is used to input the consulting question original semantic coding vector and the template question semantic coding vector into the joint implicit feature capture network to obtain a consulting question-template question semantic joint implicit feature vector; a consulting question-template question conditional feature generation secondary subunit 14312, which is used to perform feature activation based on the Sigmoid function on the consulting question-template question semantic joint implicit feature vector to obtain the consulting question-template question conditional feature vector.

[0047] More specifically, in an embodiment of the present application, the consulting question-template question semantic joint secondary sub-unit is used to: after adding the original semantic coding vector of the consulting question and the semantic coding vector of the template question by position, multiply the obtained consulting question-template question semantic sum vector by the weight matrix and then add the vector by position with the bias vector to obtain the consulting question-template question semantic joint interaction vector; use the tanh function to process the consulting question-template question semantic joint interaction vector to obtain the consulting question-template question semantic joint implicit feature vector.

[0048] Specifically, in the embodiment of the present application, the consulting question-template question feature interaction subunit 1432 includes: a consulting question semantic contribution calculation secondary subunit, which is used to calculate the consulting question semantic contribution of the original semantic coding vector of the consulting question relative to the consulting question-template question conditional feature vector; a template question semantic contribution calculation secondary subunit, which is used to calculate the template question semantic contribution of the template question semantic coding vector relative to the consulting question-template question conditional feature vector; a consulting template question semantic modulation secondary subunit, which is used to normalize the consulting question semantic contribution and the template question semantic contribution, and use the normalized consulting question semantic contribution and the normalized template question semantic contribution to modulate the consulting question semantic contribution. The original semantic coding vector of the consulting question and the semantic coding vector of the template question are weighted modulated to obtain the modulated original semantic coding vector of the consulting question and the modulated semantic coding vector of the template question; the consulting question enhanced secondary subunit is used to use the modulated original semantic coding vector of the consulting question as the query vector, the modulated template question semantic coding vector as the key vector and the consulting question-template question conditional feature vector as the value vector, and input the modulated original semantic coding vector of the consulting question, the modulated template question semantic coding vector and the consulting question-template question conditional feature vector into the feature-significant guided interaction module based on the converter structure to obtain the consulting question enhanced semantic coding vector as the consulting question enhanced semantic representation.

[0049] More specifically, in an embodiment of the present application, the second-level sub-unit for calculating the semantic contribution of the consulting question is used to: calculate the positional division of the original semantic coding vector of the consulting question and the consulting question-template question conditional feature vector to obtain the consulting question semantic contribution vector; calculate the base-two logarithmic function value of the absolute value of each eigenvalue of the consulting question semantic contribution vector to obtain the consulting question semantic contribution logarithmic vector; calculate the positional dot product of the original semantic coding vector of the consulting question and the consulting question semantic contribution logarithmic vector, and perform position-by-position multiplication on the obtained dot product vector to obtain the consulting question semantic contribution value; calculate the exponential function of the consulting question semantic contribution value with the natural constant e as the base to obtain the consulting question semantic contribution. Similarly, the processing method of the template question semantic contribution is consistent with the processing method of the consulting question semantic contribution.

[0050] More specifically, in an embodiment of the present application, the consulting template question semantic modulation secondary sub-unit is used to: calculate the sum of the consulting question semantic contribution and the template question semantic contribution to obtain the consulting question-template question semantic contribution sum; respectively divide the consulting question semantic contribution and the template question semantic contribution by the consulting question-template question semantic contribution sum to obtain the normalized consulting question semantic contribution and the normalized template question semantic contribution; perform positional point multiplication on the consulting question original semantic coding vector and the normalized consulting question semantic contribution to obtain the modulated consulting question original semantic coding vector; perform positional point multiplication on the template question semantic coding vector and the normalized template question semantic contribution to obtain the modulated template question semantic coding vector.

[0051] More specifically, in an embodiment of the present application, the consulting question enhancement secondary sub-unit is used to: perform vector multiplication of the modulated consulting question original semantic coding vector and the transposed vector of the modulated template question semantic coding vector, and perform positional division of the obtained modulated consulting question-template question semantic association matrix and the square root of the length of the modulated template question semantic coding vector to obtain a modulated consulting question-template question semantic association scaling matrix; input the modulated consulting question-template question semantic association scaling matrix into a softmax function, and perform matrix-vector multiplication of the obtained modulated consulting question-template question semantic association scaling activation matrix and the consulting question-template question conditional feature vector to obtain the consulting question enhanced semantic coding vector.

[0052] In the embodiment of the present application, specifically, the significant interaction between the features of the conditionally guided original semantic encoding vector of the consulting question and the semantic encoding vector of the template question can be expressed by the formula:

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061]

[0062] in, and are the original semantic coding vector of the consulting question and the semantic coding vector of the template question, respectively. To add by location, and are the weight matrix and bias vector respectively, for function, is the implicit feature vector of the consulting question-template question semantic union, for function, represents the consulting question-template question condition feature vector, is the feature value of each position in the original semantic encoding vector of the consulting question, is the feature value of each position in the template question semantic encoding vector, is the eigenvalue of each position in the consulting question-template question conditional eigenvector, is the number of eigenvalues ​​in the original semantic encoding vector of the consulting question and the semantic encoding vector of the template question, represents the logarithmic function value with base 2, represents an exponential function with the natural constant e as the base, The semantic contribution of the consulting question, is the semantic contribution of the template question, and They are the normalized semantic contribution of consulting questions and the normalized semantic contribution of template questions, and They are the original semantic coding vector of the consultation question after modulation and the semantic coding vector of the template question after modulation, yes The transposed vector of yes Length, represents vector multiplication, yes function, is the consulting question enhanced semantic encoding vector.

[0063] In an embodiment of the present application, the problem body index splicing enhancement module 144 is used to embed the user's body index data, and then splice the obtained body index global embedded semantic features and the consultation question enhanced semantic representation to obtain the consultation question-body index splicing semantic representation. Specifically, in an embodiment of the present application, the problem body index splicing enhancement module is used to: embed the user's body index data using the body index embedding coding matrix to obtain the body index global embedded semantic coding vector as the body index global embedded semantic feature; splice the consultation question enhanced semantic coding vector and the body index global embedded semantic coding vector to obtain the consultation question-body index splicing semantic coding vector as the consultation question-body index splicing semantic representation. It should be understood that, considering that the user's body index data contains information of multiple dimensions, such as blood pressure, blood sugar, heart rate, etc., and in order to convert the body index data into a vector representation with semantic meaning, in the technical solution of the present application, the body index embedding coding matrix is ​​used to embed the user's body index data to obtain the body index global embedded semantic coding vector. That is, the relationship and mutual influence between indicators can be reflected through the body indicator embedding coding matrix, so as to better capture the intrinsic meaning of the data. Accordingly, in order to integrate the semantic information of the user's consulting questions with the physiological information of the user's body indicators, thereby providing a more comprehensive semantic representation, in the technical solution of the present application, the consulting question enhanced semantic coding vector and the body indicator global embedded semantic coding vector are spliced ​​to obtain a consulting question-body indicator spliced ​​semantic coding vector. In this way, the user's body indicator data can be combined with the semantic information of the consulting question, so that the system can consider the user's physiological condition and specific problem description at the same time, thereby providing a more comprehensive and accurate consulting answer text.

[0064] In an embodiment of the present application, the consulting question text generation module 145 is used to obtain the consulting answer text based on the consulting question-physical index splicing semantic representation. Specifically, in an embodiment of the present application, the consulting question text generation module is used to: input the consulting question-physical index splicing semantic coding vector into the intelligent consulting module based on the large language model to obtain the consulting answer text. That is, the consulting question-physical index splicing semantic representation obtained by splicing the consulting question enhanced semantic coding vector and the physical index global embedded semantic coding vector is used to generate, so as to intelligently obtain the consulting answer text. In particular, in a specific embodiment of the present application, the large language model may include BERT, GPT series or other similar models, so that personalized consulting answer text can be generated based on the user's specific questions and physical index data, and accurate and timely suggestions can be provided. In this way, the consulting answer text can be generated instantly, reducing the user's waiting time, and more personalized health advice can be provided at the same time, assisting the doctor's medical and health consulting work, reducing the doctor's workload, and providing more personalized consulting services.

[0065] It should be understood that in the technical solution of the present application, the original semantic coding vector of the consulting question and the semantic coding vector of the template question represent the text semantic embedding coding features of the text description of the consulting question input by the user and the semantic embedding coding features of the template question, respectively. Considering that the expression of the template question is more standardized and neat compared to the text description of the consulting question input by the user, this makes the template question semantic coding vector have a relatively better feature manifold detail expression than the original semantic coding vector of the consulting question in the semantic embedding space. However, this high-dimensional feature manifold expression difference will cause the extraction of the joint implicit features between the original semantic coding vector of the consulting question and the template question semantic coding vector to be fine-grained distorted when the features between the original semantic coding vector of the consulting question and the template question semantic coding vector are conditionally guided to interact significantly, which leads to the feature manifold of the enhanced semantic coding vector of the consulting question obtained through significant guided interaction between features having expression sparsity and uneven distribution in the high-dimensional feature space.

[0066] Based on this, in a preferred embodiment, before the original semantic coding vector of the consulting question and the semantic coding vector of the template question are input into the inter-feature significant guided interaction module with the joint implicit feature as conditional information, the original semantic coding vector of the consulting question is subjected to feature manifold optimization to obtain an optimized original semantic coding vector of the consulting question, specifically comprising: performing L random sampling on the original semantic coding vector of the consulting question to obtain L consulting question semantic random perturbation eigenvalues; arranging the L consulting question semantic random perturbation eigenvalues ​​in order from large to small to obtain a consulting question semantic random perturbation ordered vector; calculating the original semantic coding vector of the consulting question The inner product of the code vector is performed to obtain the semantic energy modulation eigenvalue of the consulting question; based on the semantic energy modulation eigenvalue of the consulting question semantics, the semantic intensity of the randomly perturbed ordered vector of the consulting question semantics is scaled to obtain the semantically enhanced randomly perturbed ordered vector of the consulting question semantics; the product of the semantically enhanced randomly perturbed ordered vector of the consulting question semantics and the transposed vector of the original semantic coding vector of the consulting question is calculated to obtain the orthogonal modulation matrix of the best sample prompt of the consulting question semantics; the matrix product of the orthogonal modulation matrix of the best sample prompt of the consulting question semantics and the original semantic coding vector of the consulting question is calculated to obtain the optimized original semantic coding vector of the consulting question semantics.

[0067] The above optimization process is expressed as follows:

[0068]

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075]

[0076] in, and Indicates the original semantic encoding vector of the consulting question The random sampling The semantic random perturbation eigenvalues ​​of consulting questions, represents L random samplings, represents the Gaussian density probability function, represents the mean of the original semantic encoding vector of the consulting question, represents the original semantic encoding vector of the consulting question, It means to calculate the mean of the vector. represents the variance of the original semantic encoding vector of the consulting question, is the first The eigenvalues ​​at the positions, is the length of the original semantic encoding vector of the consulting question, Express Arrange them in order from large to small. represents the semantic random perturbation ordered vector of the consulting question, represents the square of the vector's two-norm, represents the semantic energy modulation feature value of the consulting question, It means point multiplication by position. represents matrix multiplication, represents the transposed vector of the original semantic encoding vector of the consulting question, The best example of the semantics of the consulting question indicates the orthogonal modulation matrix, Represents the optimized original semantic encoding vector of the consulting question.

[0077] Accordingly, by randomly sampling the original semantic encoding vector of the consulting question L times to obtain a set of new feature values ​​based on the original features but containing slight changes, randomness is introduced to explore different aspects of the feature space, and by simulating the noise or variation in the actual data, it helps the model to better generalize to unseen data. Then, these consulting question semantic random perturbation feature values ​​are arranged in order from large to small to form a consulting question semantic random perturbation ordered vector to identify which perturbation directions may be the most important for the current problem. Such an ordering not only helps the algorithm focus on the feature differences that can best distinguish different categories, but also helps to reveal hidden data structures or patterns. Subsequently, the inner product of the original semantic encoding vector of the consulting question is calculated to obtain the consulting question semantic energy modulation feature value, which reflects the overall strength or "energy" of the original feature vector and provides a benchmark for measuring the importance of the feature. Using this consulting question semantic energy modulation feature value as a reference point, the previously formed consulting question semantic random perturbation ordered vector is semantically scaled to construct a semantically enhanced consulting question semantic random perturbation ordered vector. In this way, those perturbations that appear to be more important in the original feature space are amplified, enhancing the model's sensitivity to key feature changes while retaining relatively small but potentially still useful perturbation information. Further, the product of the semantically enhanced consulting question semantic random perturbation ordered vector and the transpose of the original semantic encoding vector of the consulting question is calculated to obtain the consulting question semantic best example prompt orthogonal modulation matrix, the purpose of which is to find the way in which features interact from two different perspectives. The resulting matrix can be regarded as a conversion tool for capturing and representing complex linear relationships between features. Finally, by applying the above-constructed consulting question semantic best example prompt orthogonal modulation matrix to the original semantic encoding vector of the original consulting question, the final feature structure modulation is completed to make the features more in line with the requirements of the objective function, reduce redundant information, and maximize the retention of useful information, so as to improve learning efficiency and prediction performance.

[0078] In summary, the feature manifold optimization process of the original semantic coding vector of the consulting question is clarified, which extracts the best prompt examples based on fine-grained target retrieval with random perturbations, and then guides adversarial robustness learning based on the best prompt examples to complete the structure of the high-dimensional manifold of the feature vector in the feature space, so as to improve the detail and structure of the feature expression. In this way, the detail expression of the feature manifold of the original semantic coding vector of the consulting question is improved, thereby improving the fine-grained expression of the significant guided interaction encoding between the original semantic coding vector of the consulting question and the template question semantic coding vector based on the joint implicit feature as the conditional information. In this way, the consultation answer text can be generated instantly, reducing the user's waiting time, and at the same time, more personalized health advice can be provided to assist doctors in their medical and health consultation work, reduce the workload of doctors, and provide more personalized consulting services.

[0079] Therefore, the intelligent question-and-answer terminal 140 is explained, which obtains the user's body indicator data collected by the user detection terminal, and obtains the consultation question text description entered by the user, and adopts natural language processing technology and a large language model to first find a template question that matches the consultation question text description, and then semantically encodes the consultation question text description and the template question, and then conditionally guides the interaction between the original semantic features of the consultation question and the semantic features of the template question to obtain the enhanced semantic features of the consultation question, and at the same time globally embeds the user's body indicator data to intelligently obtain the consultation answer text based on the fusion semantics between the global embedded semantic features of the body indicators. In this way, the consultation answer text can be generated instantly, reducing the user's waiting time, and at the same time, it can provide more personalized health advice, assist doctors in medical and health consultation work, reduce the workload of doctors, and provide more personalized consulting services.

[0080] In summary, the telemedicine health consultation and management platform 100 based on the embodiment of the present application is explained, which includes: a user detection terminal, a user terminal, a health management platform background server and an intelligent question and answer terminal. First, the body index data is collected through the user detection terminal, and the user terminal receives the text description of the consultation question. This information is stored and managed by the health management platform background server, and then the intelligent question and answer terminal uses the collected data to provide users with customized answers based on their personal health status and specific questions, thereby realizing the instant generation of consultation answer texts and providing more personalized consultation services.

[0081] As described above, the telemedicine health consultation and management platform 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a telemedicine health consultation and management algorithm. In one possible implementation, the telemedicine health consultation and management platform 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the telemedicine health consultation and management platform 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the telemedicine health consultation and management platform 100 can also be one of the many hardware modules of the wireless terminal.

[0082] Alternatively, in another example, the remote medical health consultation and management platform 100 and the wireless terminal may also be separate devices, and the remote medical health consultation and management platform 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0083] The above descriptions of various implementations of the present disclosure are exemplary and not exhaustive. The disclosure is not limited to the disclosed implementations, and many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The terms used herein are selected to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various practical methods disclosed herein.

Claims

1. A remote medical health consultation and management platform, comprising: User detection terminal, used to collect data on various indicators of the user's body; A user terminal is used to receive a text description of a consultation question input by a user; The health management platform backend server is used to store user information; the remote medical health consultation and management platform also includes: an intelligent question and answer terminal, which is used to obtain a consultation answer text based on the consultation question text description input by the user and the user's physical index data; wherein the intelligent question and answer terminal includes: A user information acquisition module, used to acquire the user's physical index data collected by the user detection terminal, and acquire the text description of the consultation question input by the user; A text description semantic encoding module, used for obtaining a template question matching the text description of the consulting question based on the text description of the consulting question, and performing semantic encoding on the text description of the consulting question and the template question matching the text description of the consulting question to obtain original semantic features of the consulting question and semantic features of the template question; A consulting question enhancement module, used for performing conditionally guided inter-feature significant interaction on the original semantic features of the consulting question and the semantic features of the template question to obtain an enhanced semantic representation of the consulting question; A question body index splicing enhancement module is used to embed and encode the user's body index data, and then splice the obtained body index global embedding semantic features and the consultation question enhanced semantic representation to obtain a consultation question-body index splicing semantic representation; A consultation question text generation module, used to obtain the consultation answer text based on the consultation question-physical index splicing semantic representation; Wherein, the consultation question enhancement module includes: A consulting question-template question latent variable feature combination subunit is used to combine the original semantic encoding vector of the consulting question with the semantic encoding vector of the template question to obtain a consulting question-template question conditional feature vector; A consulting question-template question feature interaction subunit, used for performing conditionally dependent feature interaction fusion on the original semantic encoding vector of the consulting question and the semantic encoding vector of the template question based on the consulting question-template question conditional feature vector to obtain an enhanced semantic representation of the consulting question; The consulting question-template question latent variable feature joint subunit includes: A consulting question-template question semantic joint secondary subunit, used for inputting the consulting question original semantic encoding vector and the template question semantic encoding vector into a joint implicit feature capture network to obtain a consulting question-template question semantic joint implicit feature vector; A consulting question-template question conditional feature generation secondary subunit is used to perform feature activation based on a Sigmoid function on the consulting question-template question semantic joint implicit feature vector to obtain the consulting question-template question conditional feature vector; The consulting question-template question semantic joint secondary sub-unit is used for: After adding the original semantic coding vector of the consulting question and the semantic coding vector of the template question by position, the obtained consulting question-template question semantic sum vector is multiplied by the weight matrix and then added by position with the bias vector to obtain the consulting question-template question semantic joint interaction vector; Using a tanh function to process the consulting question-template question semantic joint interaction vector to obtain the consulting question-template question semantic joint implicit feature vector; Wherein, the consulting question-template question feature interaction subunit includes: A consulting question semantic contribution calculation secondary subunit is used to calculate the consulting question semantic contribution of the original semantic encoding vector of the consulting question relative to the consulting question-template question conditional feature vector; A template question semantic contribution calculation secondary subunit is used to calculate the template question semantic contribution of the template question semantic encoding vector relative to the consulting question-template question conditional feature vector; The consultation template question semantic modulation secondary sub-unit is used to normalize the consultation question semantic contribution and the template question semantic contribution, and use the normalized consultation question semantic contribution and the normalized template question semantic contribution to perform weighted modulation on the consultation question original semantic coding vector and the template question semantic coding vector to obtain the modulated consultation question original semantic coding vector and the modulated template question semantic coding vector; A consulting question enhancement secondary subunit is used to use the modulated original semantic coding vector of the consulting question as a query vector, the modulated template question semantic coding vector as a key vector and the consulting question-template question conditional feature vector as a value vector, and input the modulated original semantic coding vector of the consulting question, the modulated template question semantic coding vector and the consulting question-template question conditional feature vector into a feature-to-feature significant guided interaction module based on a converter structure to obtain a consulting question enhanced semantic coding vector as the consulting question enhanced semantic representation; The consultation question semantic contribution calculation secondary sub-unit is used to: Calculate the positional division of the original semantic encoding vector of the consulting question and the consulting question-template question conditional feature vector to obtain a consulting question semantic contribution vector; Calculate the logarithmic function value with base 2 of the absolute value of each eigenvalue of the consultation question semantic contribution vector to obtain the consultation question semantic contribution logarithmic vector; Calculate the positional dot product of the original semantic coding vector of the consulting question and the logarithmic vector of the semantic contribution of the consulting question, and perform positional multiplication on the obtained dot product vector to obtain the semantic contribution value of the consulting question; An exponential function with the natural constant e as the base and the semantic contribution value of the consulting question as the exponent is calculated to obtain the semantic contribution degree of the consulting question.

2. The remote medical health consultation and management platform according to claim 1, characterized in that: The text description semantic encoding module is used to: Based on the text description of the consulting question, extracting a template question matching the text description of the consulting question from a template question database; The text description of the consulting question and the template question matching the text description of the consulting question are semantically encoded to obtain an original semantic encoding vector of the consulting question as the original semantic feature of the consulting question and a semantic encoding vector of the template question as the semantic feature of the template question.

3. The remote medical health consultation and management platform according to claim 2 is characterized in that: The consultation template question semantic modulation secondary sub-unit is used to: Calculating the sum of the semantic contribution of the consulting question and the semantic contribution of the template question to obtain a sum of the semantic contributions of the consulting question and the template question; Respectively dividing the consulting question semantic contribution and the template question semantic contribution by the consulting question-template question semantic contribution sum to obtain the normalized consulting question semantic contribution and the normalized template question semantic contribution; Multiplying the original semantic coding vector of the consulting question by the normalized semantic contribution of the consulting question by position point to obtain the modulated original semantic coding vector of the consulting question; The template question semantic coding vector and the normalized template question semantic contribution are multiplied point by point to obtain the modulated template question semantic coding vector.

4. The remote medical health consultation and management platform according to claim 3 is characterized in that: The consulting question enhances the secondary subunit to: Performing vector multiplication of the original semantic coding vector of the post-modulation consulting question and the transposed vector of the semantic coding vector of the post-modulation template question, dividing the obtained post-modulation consulting question-template question semantic association matrix by the square root of the length of the semantic coding vector of the post-modulation template question by position to obtain a post-modulation consulting question-template question semantic association scaling matrix; The modulated consulting question-template question semantic association scaling matrix is ​​input into the softmax function, and the obtained modulated consulting question-template question semantic association scaling activation matrix is ​​matrix-vector multiplied with the consulting question-template question conditional feature vector to obtain the consulting question enhanced semantic coding vector.

5. The remote medical health consultation and management platform according to claim 4, characterized in that: The problem body index splicing enhancement module is used to: Use a body index embedding coding matrix to embed the user's body index data to obtain a body index global embedding semantic coding vector as the body index global embedding semantic feature; The consulting question enhanced semantic coding vector and the body indicator global embedded semantic coding vector are spliced ​​to obtain a consulting question-body indicator splicing semantic coding vector as the consulting question-body indicator splicing semantic representation.

6. The remote medical health consultation and management platform according to claim 5, characterized in that: The consultation question text generation module is used to: input the consultation question-physical index splicing semantic coding vector into the intelligent consultation module based on the large language model to obtain the consultation answer text.

Images