Health consultation shunting method and device, equipment and storage medium

Abstract

The invention provides a health consultation shunting method and device, equipment and a storage medium, and relates to the technical field of internet medical treatment. The method comprises the following steps: acquiring health consultation chief complaint information of a user; encoding the health consultation chief complaint information to obtain a feature vector of the user; processing the feature vector of the user based on a preset prediction model to obtain an initial service type of the user, the preset prediction model including at least two of an entity matching model, a deep learning classification model and a regular model; and determining a target service type of the user based on a preset screening rule and the initial service type. The shunting result is closer to the actual demand of an online platform, and the multi-model system reduces the degree of dependence on data annotation, so that the user experience is improved, and the platform and user cost is reduced.

Description

Background Art

[0002] In recent years, with the rapid development of the Internet medical industry, the scale of the online medical market has been growing day by day, which has promoted the cultivation of users' online drug purchase and medical consultation mentality. Users can submit health consultation needs about products on the product details page, such as doctor consultation, pharmacist consultation or health management consultation, etc. Users find it difficult to judge the boundaries of various types of consultation and cannot accurately determine the target consultation project.

[0003] In related technologies, subjective diversion solutions based on product interaction forms are often adopted. However, the above subjective diversion solutions have problems such as low diversion accuracy and efficiency, high cost, and professional risks.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The present disclosure provides a health consultation diversion, device, equipment and storage medium, which at least to a certain extent overcomes the problems of low efficiency, high cost and professional risks of health consultation diversion in related technologies.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0007] According to one aspect of the present disclosure, a health consultation diversion method is provided, including: obtaining a user's health consultation chief complaint information; encoding the health consultation chief complaint information to obtain a feature vector of the user; processing the user's feature vector based on a preset prediction model to obtain an initial service type for the user, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model; determining a target service type for the user based on preset screening rules and the initial service type, and sending the health consultation chief complaint information to a task queue corresponding to the target service type.

[0008] In one embodiment of the present disclosure, the preset prediction model includes an entity matching model, a deep learning classification model and a regular model, and the feature vector includes a first feature vector and a second feature vector, wherein the first feature vector includes the user's main complaint entity and the corresponding entity type; wherein the processing of the user's feature vector based on the preset prediction model to obtain the user's initial service type includes: processing the first feature vector based on the preset entity matching model to obtain the first initial service type; processing the first feature vector based on the preset regular model to obtain the second initial service type; processing the second feature vector based on the preset deep learning classification model to obtain the third initial service type; determining the user's target service type based on the preset screening rules and the initial service type includes: determining the user's target service type based on the preset screening rules, the first initial service type, the second initial service type and the third initial service type.

[0009] In one embodiment of the present disclosure, the first feature vector is processed based on a preset entity matching model to obtain a first initial service type, including: based on a preset entity service correspondence, searching for a process service type that matches the first feature vector, wherein the preset entity service correspondence includes a correspondence between a main complaint entity, an entity type, and a service type; if the number of the process service types is one, taking the target service type as the first initial service type; if the number of the process service types is multiple, determining the first initial service type based on the process service types and voting weights.

[0010] In one embodiment of the present disclosure, the first feature vector is processed based on a preset regular model to obtain a second initial service type, including: if the first feature vector satisfies a preset regular expression, determining that the second initial service type is a customer service consulting service type.

[0011] In one embodiment of the present disclosure, the first feature vector is obtained by: encoding the health consultation chief complaint information to obtain a chief complaint vector; performing context matching on the chief complaint vector to obtain an encoded chief complaint vector; and decoding the encoded chief complaint vector to obtain the first feature vector.

[0012] In one embodiment of the present disclosure, the second feature vector is obtained by: converting the user's health consultation complaint information into pinyin text; and semantically encoding the pinyin text to obtain the second feature vector.

[0013] In one embodiment of the present disclosure, determining the target service type of the user based on preset screening rules and the initial service type includes: determining the target service type of the user based on preset decision priorities and the initial service type, wherein the preset decision priorities are, from high to low, regular model, entity matching model, and deep learning classification model.

[0014] In one embodiment of the present disclosure, determining the target service type of the user based on preset screening rules and the initial service type includes: extracting the business scenario of the user; and determining the target service type of the user based on the business scenario and the initial service type.

[0015] In one embodiment of the present disclosure, after encoding the health consultation complaint information to obtain the feature vector of the user, the method further includes: if the health consultation complaint information of the user contains potential health risk information, determining that the target service type of the user is a doctor consultation service type.

[0016] In another aspect of the present disclosure, a health consultation diversion device is also provided, including: an acquisition module for acquiring a user's health consultation chief complaint information; an encoding module for encoding the health consultation chief complaint information to obtain a feature vector of the user; a prediction module for processing the user's feature vector based on a preset prediction model to obtain an initial service type of the user, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model and a regular model; a screening module for determining the user's target service type based on preset screening rules and the initial service type, and sending the health consultation chief complaint information to a task queue corresponding to the target service type.

[0017] According to another aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the above-mentioned health consultation diversion method by executing the executable instructions.

[0018] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned health consultation diversion method is implemented.

[0019] According to another aspect of the present disclosure, a computer program product is provided, which includes a computer program or computer instructions, and the computer program or the computer instructions are loaded and executed by a processor to enable a computer to implement the above-mentioned health consultation diversion method.

[0020] In an embodiment of the present disclosure, the user's health consultation chief complaint information is obtained, encoded to obtain the user's feature vector, and the feature vector is processed based on multiple preset prediction models such as entity matching model, deep learning classification model, regularization model, etc. to obtain the user's initial service type. Based on the preset screening rules and the initial service model, the user's target service type can be determined. The present disclosure conducts model training and testing based on the real online user's chief complaint information, so that the diversion results are closer to the timing requirements of the online platform, combines multiple prediction models, reduces the dependence on data annotation, and realizes intelligent diversion of user consultation from the perspective of health professionalism, thereby improving user experience, promoting order volume, reducing platform and user costs, and reducing potential professional risks.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0023] Figure 1 The following is an architecture diagram of a health consultation and diversion system provided by an embodiment of the present disclosure;

[0024] Figure 2 A flow chart of a health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0025] Figure 3 A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0026] Figure 4 A flow chart of a method for determining a first initial service type provided by an embodiment of the present disclosure is shown;

[0027] Figure 5 A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0028] Figure 6 A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0029] Figure 7 An example flow chart of a health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0030] Figure 8A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0031] Fig. 9 A schematic diagram showing a health consultation diversion method provided by an embodiment of the present disclosure is shown;

[0032] Fig.10 A schematic diagram showing the structure of a health consultation diversion device provided by an embodiment of the present disclosure is shown;

[0033] Fig.11 A structural block diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0034] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0035] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0036] Figure 1 An exemplary system architecture 100 is shown to which the health consultation diversion method or health consultation diversion device according to the embodiments of the present disclosure can be applied.

[0037] like Figure 1 As shown, the system architecture 100 may include a terminal device 110 , a network 120 and a server 130 .

[0038] The network 120 is a medium for providing a communication link between the terminal device 110 and the server 130 , and may be a wired network or a wireless network.

[0039] The user may use the terminal device 110 to interact with the server 130 through the network 120 to receive or send messages.

[0040] The terminal device 110 may be any electronic device having a display screen and supporting web browsing, including but not limited to a smart phone, a tablet computer, a display screen, and a desktop computer.

[0041] The client of the application program that can be installed on the terminal device 110 is the same, or the client of the same type of application program based on different operating systems. Based on the different terminal platforms, the specific form of the client of the application program can also be different, for example, the application program client can be a mobile client, a PC client, etc.

[0042] Exemplarily, the terminal device 110 is installed with various communication client applications, such as medical applications, web browser applications, search applications, etc. The communication client application can collect the user's health consultation complaint information, and determine the target service type based on the collected health consultation complaint information, so as to divert the user to the consultation service corresponding to the target service type, and send the health consultation complaint information to the task queue corresponding to the target service type.

[0043] The server 130 may be a server that provides various services, such as a background management server that supports the device operated by the user using the terminal device 110. The background management server may analyze and process the received request and other data (such as health consultation complaint information, etc.) to obtain a processing result (such as the target service type, sending the health consultation complaint information to the task queue corresponding to the target service type).

[0044] Optionally, server 130 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0045] Those skilled in the art will know that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration, and any number of terminal devices, networks and servers may be provided according to actual needs, and the embodiments of the present disclosure do not limit this.

[0046] In the relevant technologies, most medical e-commerce platforms often use subjective diversion methods based on product interaction forms for user health consultation diversion solutions in the business details page scenario. For example, multiple consultation entrances such as "Ask customer service", "Ask a doctor", and "Ask a nutritionist" are designed on the business details page, and multiple sub-category consultation entrances are designed on the secondary landing page. The product interaction is progressive, and users make subjective choices layer by layer. When the user finally chooses a doctor for consultation, the department dimension is triaged based on AI triage; when the user finally chooses consulting services such as pharmacists or nutritionists, they are directly diverted to the corresponding consulting services.

[0047] However, the subjective diversion solution based on product interaction has many shortcomings, as follows:

[0048] Due to the different levels of understanding of health knowledge among users, it is difficult to define the responsibilities of customer service, doctors, pharmacists, nutritionists and other roles, which makes it easy to make mistakes when choosing services such as "ask customer service", "ask pharmacists" and "ask doctors", which has a negative impact on user experience, order volume, cost control, etc. In addition, the service type diversion method based on user subjective consciousness may have professional risks and affect the reputation of the platform.

[0049] First, the subjective diversion method based on the product interaction form affects the user experience and order volume. When a user wants to consult an open question about a prescription drug, the actual operation on the product details page chooses the "Ask the Pharmacist" service, but the pharmacist cannot solve the user's open needs and cannot transfer orders to doctors across specialties, which seriously affects the user experience and the final order transaction.

[0050] Secondly, the subjective triage method based on cost interaction results in cost waste. When a user wants to consult about the usage and dosage of an over-the-counter drug, the actual operation on the product details page selects the "Ask a doctor" service. Although doctors can solve such consultation problems, considering that the number of doctors is limited and the cost of doctor consultation is often higher, it is a cost waste for both users and the platform.

[0051] Finally, the subjective triage method based on cost interaction has professional risks. When users choose to consult a pharmacist, the health consultation information entered by the user is the description of the user's current symptoms and the questions about the usage and dosage of the medicine, and the symptoms described by the user are serious. If the pharmacist does not provide timely medical reminders during the answering process, there may be professional risks. At the same time, there is a time delay in subsequent quality control, and such risks cannot be eliminated in time.

[0052] Based on this, the present disclosure designs a health consultation diversion system and related methods to solve at least part of the health consultation diversion problems mentioned above.

[0053] The health consultation diversion method provided by the present invention obtains the user's health consultation chief complaint information, encodes the health consultation chief complaint information, obtains the user's feature vector, processes the feature vector based on multiple preset prediction models such as entity matching model, deep learning classification model, regularization model, etc., obtains the user's initial service type, and determines the user's target service type based on preset screening rules and the initial service model, and sends the health consultation chief complaint information to the task queue corresponding to the target service type. The present invention conducts model training and testing based on the real online user chief complaint information, so that the diversion result is closer to the timing requirements of the online platform, combines multiple prediction models, reduces the dependence on data annotation, realizes intelligent diversion of user consultation from the perspective of health professionalism, improves user experience, promotes order volume, reduces platform and user costs, and reduces potential professional risks.

[0054] It should be pointed out that the embodiments of the present invention and the technical features therein may be combined with each other without conflict.

[0055] The present exemplary implementation is described in detail below with reference to the accompanying drawings and embodiments.

[0056] First, a health consultation diversion method is provided in the embodiments of the present disclosure, and the method can be executed by any system with computing processing capabilities. In some embodiments, the health consultation diversion method provided in the embodiments of the present disclosure can be executed by a server; in other embodiments, the health consultation diversion method provided in the embodiments of the present disclosure can be implemented by a terminal device and a server through interaction.

[0057] Figure 2 A flow chart of a health consultation diversion method provided in an embodiment of the present disclosure is shown as follows: Figure 2 As shown, the health consultation diversion method provided in the embodiment of the present disclosure includes the following steps:

[0058] S202: Obtain the user's health consultation complaint information.

[0059] The main complaint information of health consultation is the main basis for expressing the user's consultation intention. The user can enter the main complaint information of health consultation in the form of text or voice through the business details page on the terminal device. The terminal device transmits the collected main complaint information of the user to the server so that the server processes the main complaint information of the user and obtains the target service type suitable for the user. The main complaint information of health consultation may include the matters that the user wants to consult, or the symptoms of the user, etc., without specific limitation.

[0060] S204: Encode the health consultation complaint information to obtain the user's feature vector.

[0061] In step S204, by encoding the health consultation chief complaint information, a feature vector can be obtained, and the feature vector is used as an input of a preset prediction model, so that the preset prediction model processes the feature vector and outputs an initial service type.

[0062] S206. Process the user's feature vector based on a preset prediction model to obtain the user's initial service type, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model.

[0063] The above-mentioned preset prediction model can be used to predict the user's initial service type based on the user's feature vector. The preset prediction model can be obtained by model training and testing based on the health consultation complaint information of real online users and the real service type label, so that the diversion result is more accurate. It should be noted that the preset prediction model is a trained prediction model.

[0064] The above-mentioned preset prediction model may include at least two of an entity matching model, a deep learning classification model, and a regular model, and the training and verification process of each model may be performed separately. Specifically, the above-mentioned preset prediction model may be determined based on a loss function (e.g., the difference between the initial service type obtained by the preset prediction model and the actual service type label corresponding to the health consultation chief complaint information sample).

[0065] It should be noted that the services corresponding to the initial service type may include one or more of doctor consultation services, pharmacist consultation services, health management consultation services, customer service consultation services and invalidity consultation services.

[0066] S208. Based on the preset screening rules and the initial service type, determine the user's target service type, and send the health consultation complaint information to the task queue corresponding to the target service type.

[0067] The preset screening rules in S208 are used to screen the target service type applicable to the user from the initial service type, so as to divert the user to the consulting service corresponding to the target service type, and send the health consulting complaint information to the task queue of the above consulting service. The preset screening rules can be pre-stored in the memory, and the preset screening rules can include one or more rules, for example, screening rules based on different prediction model priorities, screening rules based on application scenarios, etc.

[0068] In an embodiment of the present disclosure, the user's health consultation chief complaint information is obtained, encoded to obtain the user's feature vector, and the feature vector is processed based on multiple preset prediction models such as entity matching model, deep learning classification model, regularization model, etc. to obtain the user's initial service type. Based on the preset screening rules and the initial service model, the user's target service type can be determined, and the health consultation chief complaint information is sent to the task queue corresponding to the target service type. The present disclosure performs model training and testing based on real online user chief complaint information, so that the diversion results are closer to the timing requirements of the online platform, and combines multiple prediction models to reduce the dependence on data annotation, so as to realize intelligent diversion of user consultation from the perspective of health professionalism, improve user experience, promote order volume, reduce platform and user costs, and reduce potential professional risks.

[0069] In one embodiment, when the preset prediction model includes two prediction models, the preset prediction model may be an entity matching model and a deep learning model, or an entity matching model and a regular model, or a deep learning model and a regular model.

[0070] In one embodiment, the preset prediction model includes three prediction models, and the preset prediction model includes an entity matching model, a deep learning classification model and a regular model.

[0071] It should be noted that the implementation of the two prediction models is similar to the implementation of the three prediction models, and the implementation of the three prediction models is described below.

[0072] In one embodiment, the preset prediction model includes an entity matching model, a deep learning classification model and a regular model, and the feature vector includes a first feature vector and a second feature vector, wherein the first feature vector includes the user's main complaint entity and the corresponding entity type. The main complaint entity is used to describe the main content of the user's complaint, and the entity type is used to distinguish the type of the complaint content.

[0073] Entity types may include but are not limited to disease, Chinese medicine, Western medicine, nutrition and health care, symptoms, severe illness, nursing, surgery, examination, treatment, insult and slander, customer service, etc. Table 1 shows common entity types and their entity abbreviations.

[0074] Table 1 Common entity types and entity abbreviations

[0075] Entity Type Entity abbreviation Entity Type Entity abbreviation Entity Type Entity abbreviation disease Dis traditional Chinese medicine Cmed Western medicine Med Nutrition and Health Nut symptom Sym severe Ser Nursing Nur Operation Sur examine Exa treat Tre Insults and slander Abu customer service Cus

[0076] In one embodiment, the first feature vector is obtained by: encoding the health consultation chief complaint information to obtain a chief complaint vector; performing context matching on the chief complaint vector to obtain an encoded chief complaint vector; and decoding the encoded chief complaint vector to obtain the first feature vector.

[0077] Exemplarily, the health consultation chief complaint information is encoded using an embedding layer, and each word in the health consultation chief complaint information text is converted into a vector that can represent semantic information to obtain a chief complaint vector.

[0078] The main complaint vector is input into the LatticeLSTM model for context matching, that is, secondary encoding is performed to obtain the encoded main complaint vector.

[0079] The encoded chief complaint vector is input into the Conditional Random Field (CRF) model for decoding to obtain an entity tag, namely the first feature vector. Through the above process, it is possible to determine which characters in the health consultation chief complaint information belong to the identified chief complaint entity and the corresponding entity type.

[0080] Figure 3 Another flow chart of a health consultation diversion method according to an embodiment of the present disclosure is shown. The health consultation diversion method provided by the embodiment of the present disclosure includes:

[0081] S302: Based on a preset entity matching model, process the first feature vector to obtain a first initial service type.

[0082] The preset entity matching model is used to determine the corresponding relationship between the first feature vector and the first initial service type. The above corresponding relationship includes the corresponding relationship between the main complaint entity, the entity category, and the service type.

[0083] S304: Based on a preset regularization model, the first feature vector is processed to obtain a second initial service type.

[0084] In step S304, the method includes: if the first feature vector satisfies a preset regular expression, determining that the second initial service type is a customer service consulting service type.

[0085] In the actual implementation process, the health consultation complaint information with a high degree of freedom may not be matched to the entity matching model or deep learning classification model. In this disclosure, a regular expression is designed for consultations with fixed sentence structures, such as: pattern = (box|box|bag|grain|pack|strip|scoop|portion|jin|gram|g|Gll|L).{0,5}(more|how much|several|[0-9]). When users consult product rules, it is easy for the first feature vector to meet the requirements of the preset regular expression, thereby classifying the user into the customer service consultation service type. The form and content of the regular expression can be determined according to the actual situation.

[0086] S306. Based on a preset deep learning classification model, process the second feature vector to obtain a third initial service type.

[0087] The preset deep learning classification model can be a fully connected neural network model, which uses fully connected neural network technology to mine invisible rules from a large amount of user consultation data.

[0088] In one embodiment, the second feature vector is obtained by: converting the user's health consultation complaint information into pinyin text; and semantically encoding the pinyin text to obtain the second feature vector.

[0089] For example, the main complaint information of health consultation is converted into pinyin text to avoid the negative impact of misclassification caused by typos during free input by users. The pinyin text is input into the Bidirectional Encoder Representations from Transformers (BERT) model for semantic encoding to obtain the second feature vector. The BERT model is a deep learning pre-training model that can be trained based on the pinyin text and the corresponding labels.

[0090] The second feature vector is used as input, and after being processed by a preset deep learning classification model, codes of different consulting service types such as doctor consulting service, pharmacist consulting service, and health management consulting service are output, and the third initial service type is determined based on the above codes. The preset deep learning classification model is trained based on the second feature vector and the corresponding label.

[0091] S308: Determine the target service type of the user based on the preset screening rule, the first initial service type, the second initial service type, and the third initial service type.

[0092] In an embodiment of the present disclosure, the first feature vector is processed by an entity matching model and a regular model respectively to obtain a first initial service type and a second initial service type, and the second feature vector is processed by a deep learning classification model to obtain a third initial service type. Based on preset screening rules, the user's target service type is screened from the above initial service types. The present disclosure uses a combination of entity matching, regular matching, and deep learning to achieve health consultation classification based on professional knowledge and real data, reduce dependence on data annotation, and have a wider range of adaptability.

[0093] Figure 4 FIG. 1 is a flow chart of a method for determining a first initial service type provided by an embodiment of the present disclosure. Figure 4 As shown, the above S302 processes the first feature vector based on the preset entity matching model to obtain the first initial service type, including:

[0094] S402, searching for a process service type matching the first feature vector based on a preset entity-service correspondence relationship, wherein the preset entity-service correspondence relationship includes a correspondence relationship between a main complaint entity, an entity type, and a service type;

[0095] S404: If the number of process service types is one, use the process service type as the first initial service type.

[0096] It should be noted that the preset entity-service correspondence is used to characterize the correspondence between the main complaint entity, entity type and service type. The above correspondence can be stored in the health knowledge base of the server, which can be stored in the form of a correspondence table or in other forms such as a curve. Table 2 shows an example of entities stored in the health knowledge base.

[0097] Table 2 Examples of entities stored in the health knowledge base

[0098] Complaint Entity Entity Type Corresponding service type Vitamin C Nut Health management services Ganmaoling Granules Med Pharmacist services cough Sym Doctor Service return the goods Cus customer service Chest pain Ser Doctor Service

[0099] During the implementation process, the main complaint information of health consultation has high freedom. When matching entities, precise matching is given priority. If the main complaint entity obtained based on the main complaint information of health consultation is consistent with the main complaint entity stored in the health knowledge base, the match is determined to be successful, and the service type corresponding to the main complaint entity that is successfully matched is used as the process service type. If the main complaint entity obtained based on the main complaint information of health consultation is consistent with the main complaint entity stored in the health knowledge base, it is determined that the match is not successful, and the main complaint entity that is not successfully matched and has a length of more than three characters is fuzzy matched.

[0100] The specific process of fuzzy matching is: when the first word of the main complaint entity is consistent and the length of the main complaint entity is 4 or 5 characters, 1 edit distance is allowed; when the first word of the main complaint entity is consistent and the length of the main complaint entity is more than 5 characters, 2 edit distances are allowed. When the similarity between the main complaint entity obtained according to the main complaint information of the health consultation and the main complaint entity stored in the health knowledge base exceeds the preset score threshold (for example, 85 points), the match is determined to be successful.

[0101] It should be noted that, in the process of fuzzy matching, the values ​​of the above-mentioned character length and edit distance can be determined according to actual conditions and are not specifically limited.

[0102] Continue to refer Figure 4 The health consultation diversion method provided by the embodiment of the present disclosure also includes:

[0103] S406: If there are multiple target service types, determine a first initial service type based on the target service type and the voting weight.

[0104] After completing the entity matching, since the health consultation chief complaint information may contain multiple different types of chief complaint entities, and the service types corresponding to different types of chief complaint entities may also be different, in response to this situation, the present invention assigns different voting weights to more than one process service types based on the priority strategy, and makes voting decisions in order from high to low priority. The voting weight of high-priority entity types is higher, and the voting weight of low-priority entity types is low. Then, the first initial service type is determined according to the process service type and the voting results.

[0105] The priority of entity types configured in the health knowledge base is from high to low: serious illness, invalid complaints (insults and slander), customer service consultation (customer service), doctor consultation (diseases, symptoms, surgeries, examinations, treatments), pharmacist consultation (Western medicine, traditional Chinese medicine), and health management (nursing, nutrition and health care).

[0106] For example, if the main complaint information of a health consultation is [I caught a cold recently, and the Ganmao Ling Granules didn't work, what should I do?], the main complaint information of the health consultation contains two main complaint entities, [Cold] and [Ganmao Ling Granules]. The entity type of the main complaint entity [Cold] is Dis, which corresponds to the doctor service type, and the entity type of the main complaint entity [Ganmao Ling Granules] is Med, which corresponds to the pharmacist service type. Since Dis has a higher priority than Med, the doctor service type is given a higher voting weight, and the pharmacist service type is given a lower voting weight. The doctor service type is used as the first initial service type, and the user will eventually be recommended to the doctor service type, thereby improving professionalism.

[0107] Figure 5 A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown. Figure 1 Based on the embodiment, S208 is further refined into S2082 to determine which type of model prediction result is used as the target service type. Figure 5 As shown, the health consultation diversion method provided by the embodiment of the present disclosure includes S202-S206, and S2082. Specifically, the method includes:

[0108] S2082. Determine the target service type of the user based on preset decision priorities and the initial service type, wherein the preset decision priorities are, from high to low, regular model, entity matching model, and deep learning classification model.

[0109] It should be noted that the specific implementation method of the above S202-S206 is the same as the specific implementation method of S202-S206 in the aforementioned embodiment, and will not be repeated here.

[0110] The prediction results of the regular model, entity matching model and deep learning classification model may output different service types, that is, the first initial service model, the second initial service model and the third initial service model are not exactly the same, so the initial service types need to be integrated.

[0111] Among the three prediction models, the regular model has the highest semantic interpretability, so the regular model has the highest decision priority; the entity matching model is based on the health knowledge base for matching and has a certain degree of interpretability, so the entity matching model has a medium decision priority; the deep learning classification model has poor interpretability, so the deep learning classification model has the lowest decision priority and is used as a backup logic.

[0112] The preset decision priority can be pre-configured in the server and can be represented by numbers, symbols, etc. By setting the decision priority, the prediction results of different prediction models can be integrated to improve the applicability of the model.

[0113] Figure 6 A flow chart of another health consultation diversion method provided by an embodiment of the present disclosure is shown. Figure 1 Based on the embodiment, S208 is further refined into S2084 to determine which type of model prediction result is used as the target service type. Figure 6 As shown, the health consultation diversion method provided by the embodiment of the present disclosure includes S202-S206, and S2084. Specifically, the method includes:

[0114] S2084. Extract the user's business scenario; and determine the user's target service type based on the business scenario and the initial service type.

[0115] Different business scenarios have different requirements for system classification indicators. For example, scenario A requires a higher accuracy rate for identifying a certain type of problem, while scenario B requires a higher recall rate for identifying a certain type of problem.

[0116] For example, for a business scenario where non-healthy questions (asked to customers and invalid consultations) are filtered out on the Q&A page of a business detail page, if one of the three prediction models is met, it is determined to be a non-healthy consultation.

[0117] Exemplarily, for the pre-consultation page filtering customer consultation, if two of the three prediction models are satisfied, it is determined to be a customer service consultation service.

[0118] For example, for the recommendation of consultation-type orders in the shopping guide business scenario, if one of the three prediction models is met, it is determined to be a doctor consultation service.

[0119] Table 3 lists the identification index requirements for certain types of problems in different business scenarios and the adjustment rules for the output results.

[0120] Table 3 Adjustment rules for output results in different business scenarios

[0121]

[0122]

[0123] Figure 7 FIG. 1 is a flow chart showing an example of a health consultation diversion method provided by an embodiment of the present disclosure. Figure 7 As shown, the health consultation triage method of this example includes the following steps:

[0124] S702, the regular model, the entity matching model, and the deep learning classification model are predicted respectively, and the priorities of the three prediction models are the regular model, the entity matching model, and the deep learning classification model from high to low;

[0125] S704, making a classification result decision according to the priority;

[0126] S706, personalization rule adjustment based on business scenarios;

[0127] S708. The system outputs the target service type.

[0128] Figure 8 Another health consultation diversion method process provided by the embodiment of the present disclosure is shown. Figure 2 Based on the embodiment, S2052-S2054 are added between S204 and S206 to limit the situation where the first feature vector contains potential health risk information. Figure 8 As shown, the health consultation diversion method provided by the embodiment of the present disclosure includes:

[0129] S2052, determining whether the first feature vector contains potential health risk information, if so, executing S2054; if not, executing S206;

[0130] S2054: Determine that the user's target service type is a doctor consultation service type.

[0131] It should be noted that the specific implementation method of the above S202-S208 is the same as the specific implementation method of S202-S208 in the aforementioned embodiment, and will not be repeated here.

[0132] The above-mentioned potential health risk information is information related to serious illness, such as chest pain, heart pain, etc.

[0133] In one embodiment, considering that the serious illness in the first feature vector mentioned poses a health risk to the patient, once the entity matching model matches the information related to the serious illness, the prediction results of other prediction models will not be considered, and the target service type will be output as the doctor consultation service type, so as to promptly remind the user of the health risks, consult a doctor in time, and avoid potential health risks.

[0134] In order to deepen the understanding of the implementation process of the present disclosure, the following Fig. 9 Provide detailed explanation.

[0135] like Fig. 9 As shown, the health consultation triage method of this example includes the following processes:

[0136] S902, the user inputs health consultation complaint information through the terminal device;

[0137] S906, entity recognition: Encode the health consultation complaint information to obtain a feature vector; the feature vector includes a first feature vector and a second feature vector; the first feature vector includes the user's complaint entity and the corresponding entity type;

[0138] Performing entity matching and type decision on the first feature vector based on an entity matching model to obtain a first initial service type, wherein the entity matching model performs entity matching based on entity-service correspondences stored in a health knowledge base reviewed by experts;

[0139] Processing the first feature vector based on the regular model to obtain a second initial service type;

[0140] Based on a deep learning classification model, the second feature vector is processed to obtain a third processing type; the deep learning classification model is trained by expert-labeled data;

[0141] S908: Based on the preset screening rules, the first initial service type, the second initial service type, and the third initial service type are selected as models to determine the target service model. The target service type includes but is not limited to pharmacist consultation service type, health management consultation service type, doctor consultation service type, customer service consultation service type, invalid consultation, etc.

[0142] S910, output the target service model, and send the health consultation complaint information to the task queue corresponding to the target service type.

[0143] The health consultation diversion method provided in the embodiment of the present disclosure utilizes the knowledge graph and combines the deep learning method based on the main complaint information of online health consultation to design a user health consultation intelligent diversion system to solve the health consultation diversion problem on the business details page. The system has the following technical advantages and application advantages:

[0144] In terms of technology, this system uses the health consultation complaint information of real online users for model training and testing. The diversion results are closer to the actual needs of the online platform, improving data validity. Compared with a single deep learning or knowledge graph matching implementation strategy, this system combines three strategies, namely entity matching model, regular model, and deep learning classification model, to reduce the dependence on data labeling and can handle situations where there is no corresponding entity in the knowledge graph.

[0145] In terms of application, it can realize intelligent diversion of user consultations from the perspective of health professionalism, avoid situations where user consultations cannot be answered, thereby improving user experience and increasing order volume; the system can avoid excessive answers, achieve the most professional consulting service effect at the lowest cost, and reduce platform and user costs; when the user intends to choose non-medical consulting services (pharmacist consultation or health management consultation), the system will automatically determine whether there are serious illness words in the user's chief complaint information. If so, it will recommend the user to choose doctor consultation services to avoid potential health risks and reduce potential professional risks.

[0146] Based on the same inventive concept, the embodiments of the present disclosure also provide a health consultation diversion device, such as the following embodiments. Since the principle of solving the problem in the device embodiment is similar to that in the above method embodiment, the implementation of the device embodiment can refer to the implementation of the above method embodiment, and the repeated parts will not be repeated.

[0147] Fig.10 FIG. 1 is a schematic diagram of a health consultation diversion device provided by an embodiment of the present disclosure. Fig.10 As shown, the health consultation diversion device of this embodiment includes an acquisition module 1010, an encoding module 1020, a prediction module 1020 and a screening module 1040.

[0148] Wherein, the acquisition module 1010 is used to obtain the user's health consultation complaint information;

[0149] The encoding module 1020 is used to encode the health consultation complaint information to obtain the user's feature vector;

[0150] A prediction module 1030 is used to process the user's feature vector based on a preset prediction model to obtain the user's initial service type, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model;

[0151] The screening module 1040 is used to determine the user's target service type based on preset screening rules and the initial service type, and send the health consultation complaint information to the task queue corresponding to the target service type.

[0152] It should be noted that the preset prediction model includes an entity matching model, a deep learning classification model and a regular model, and the feature vector includes a first feature vector and a second feature vector, wherein the first feature vector includes the user's main complaint entity and the corresponding entity type.

[0153] In one embodiment, the prediction module 1030 includes a first prediction module, a second prediction module and a third prediction module which are not shown in the accompanying drawings, wherein the first prediction module is used to process the first feature vector based on a preset entity matching model to obtain a first initial service type; the second prediction module is used to process the first feature vector based on a preset regular model to obtain a second initial service type; the third prediction module is used to process the second feature vector based on a preset deep learning classification model to obtain a third initial service type; the screening module 1040 is used to determine the user's target service type based on preset screening rules, the first initial service type, the second initial service type and the third initial service type.

[0154] In one embodiment, the first prediction module is specifically used to find the process service type that matches the first feature vector based on a preset entity service correspondence, wherein the preset entity service correspondence includes the correspondence between the main complaint entity, the entity type and the service type; if the number of process service types is one, the process service type is used as the first initial service type; if the number of process service types is multiple, the first initial service type is determined based on the process service type and the voting weight.

[0155] In one embodiment, the second prediction module is used to determine that the second initial service type is a customer service consulting service type if the first feature vector satisfies a preset regular expression.

[0156] It should be noted that the encoding module 1020 is used to obtain the first feature vector in the following manner: encoding the health consultation chief complaint information to obtain the chief complaint vector; performing context matching on the chief complaint vector to obtain the encoded chief complaint vector; and decoding the encoded chief complaint vector to obtain the first feature vector.

[0157] In one embodiment, the encoding module 1020 is used to obtain the second feature vector in the following manner: converting the user's health consultation complaint information into pinyin text; and semantically encoding the pinyin text to obtain the second feature vector.

[0158] In one embodiment, the screening module 1040 is used to determine the target service type of the user based on a preset decision priority and an initial service type, wherein the preset decision priorities are, from high to low, a regular model, an entity matching model, and a deep learning classification model.

[0159] In one embodiment, the screening module 1040 is used to extract the business scenario of the user; and determine the target service type of the user based on the business scenario and the initial service type.

[0160] In one embodiment, the first prediction module is further used to determine that the user's target service type is a doctor consultation service type if there is potential health risk information in the first feature vector.

[0161] In an embodiment of the present disclosure, the user's health consultation chief complaint information is obtained, encoded to obtain the user's feature vector, and the feature vector is processed based on multiple preset prediction models such as entity matching model, deep learning classification model, regularization model, etc. to obtain the user's initial service type. Based on the preset screening rules and the initial service model, the user's target service type can be determined. The present disclosure conducts model training and testing based on the real online user's chief complaint information, so that the diversion results are closer to the timing requirements of the online platform, combines multiple prediction models, reduces the dependence on data annotation, and realizes intelligent diversion of user consultation from the perspective of health professionalism, thereby improving user experience, promoting order volume, reducing platform and user costs, and reducing potential professional risks.

[0162] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as a "circuit", "module" or "system".

[0163] Refer to the following Fig.11 The electronic device 1100 according to this embodiment of the present invention is described. Fig.11 The electronic device 1100 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0164] like Fig.11 As shown, the electronic device 1100 is in the form of a general computing device. The components of the electronic device 1100 may include but are not limited to: at least one processing unit 1110, at least one storage unit 1120, and a bus 1130 connecting different system components (including the storage unit 1120 and the processing unit 1110).

[0165] The storage unit stores program codes, which can be executed by the processing unit 1110, so that the processing unit 1110 executes the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification. For example, the processing unit 1110 can execute the following steps: Figure 2As shown in , the user's health consultation chief complaint information is obtained; the health consultation chief complaint information is encoded to obtain the user's feature vector; the user's feature vector is processed based on a preset prediction model to obtain the user's initial service type, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model; based on preset screening rules and the initial service type, the user's target service type is determined, and the health consultation chief complaint information is sent to the task queue corresponding to the target service type.

[0166] The storage unit 1120 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 11201 and / or a cache storage unit 11202 , and may further include a read-only storage unit (ROM) 11203 .

[0167] The storage unit 1120 may also include a program / utility 11204 having a set (at least one) of program modules 11205, such program modules 11205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0168] Bus 1130 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0169] The electronic device 1100 may also communicate with one or more external devices 1140 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the system, and / or any device that enables the electronic device 1100 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1150. Furthermore, the system may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1160. Fig.11 As shown, the network adapter 1160 communicates with other modules of the electronic device 1100 via the bus 1130. It should be understood that although Fig.11 Not shown, other hardware and / or software modules may be used in conjunction with electronic device 1100, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0170] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0171] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, which may be a readable signal medium or a readable storage medium. In an exemplary embodiment of the present disclosure, a computer program product is also provided, which includes a computer program or a computer instruction, and the computer program or the computer instruction is loaded and executed by a processor to enable the computer to implement the steps of the method disclosed in the above embodiment.

[0172] More specific examples of computer-readable storage media in the present disclosure may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0173] In the present disclosure, a computer readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, wherein a readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A readable signal medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0174] Alternatively, the program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0175] In a specific implementation, the program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0176] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0177] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0178] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0179] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. A health consultation diversion method, characterized in that: include: Obtain the user's health consultation complaint information; Encoding the health consultation complaint information to obtain a feature vector of the user; Processing the feature vector of the user based on a preset prediction model to obtain an initial service type of the user, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model; Based on the preset screening rules and the initial service type, the target service type of the user is determined, and the health consultation complaint information is sent to the task queue corresponding to the target service type.

2. The health consultation diversion method according to claim 1, characterized in that: The preset prediction model includes an entity matching model, a deep learning classification model and a regular model, and the feature vector includes a first feature vector and a second feature vector, wherein the first feature vector includes the user's main complaint entity and the corresponding entity type; The processing of the feature vector of the user based on the preset prediction model to obtain the initial service type of the user includes: Based on a preset entity matching model, the first feature vector is processed to obtain a first initial service type; Based on a preset regular model, the first feature vector is processed to obtain a second initial service type; Based on a preset deep learning classification model, processing the second feature vector to obtain a third initial service type; The determining the target service type of the user based on the preset screening rule and the initial service type includes: Based on the preset screening rule, the first initial service type, the second initial service type, and the third initial service type, a target service type of the user is determined.

3. The health consultation diversion method according to claim 2, characterized in that: The processing of the first feature vector based on a preset entity matching model to obtain a first initial service type includes: Based on a preset entity-service correspondence, searching for a process service type that matches the first feature vector, wherein the preset entity-service correspondence includes a correspondence between a main complaint entity, an entity type, and a service type; If the number of the process service type is one, the process service type is used as the first initial service type; If there are multiple process service types, a first initial service type is determined based on the process service types and voting weights.

4. The health consultation diversion method according to claim 2, characterized in that: The processing of the first feature vector based on a preset regular model to obtain a second initial service type includes: If the first feature vector satisfies a preset regular expression, the second initial service type is determined to be a customer service consulting service type.

5. The health consultation diversion method according to claim 2, characterized in that: The first eigenvector is obtained by: Encoding the health consultation chief complaint information to obtain a chief complaint vector; Performing context matching on the main complaint vector to obtain an encoded main complaint vector; The encoded main complaint vector is decoded to obtain the first feature vector.

6. The health consultation diversion method according to claim 2, characterized in that: The second eigenvector is obtained by: Converting the user's health consultation complaint information into pinyin text; The pinyin text is semantically encoded to obtain the second feature vector.

7. The health consultation diversion method according to claim 2, characterized in that: The determining the target service type of the user based on the preset screening rule and the initial service type includes: Based on preset decision priorities and the initial service type, the target service type of the user is determined, wherein the preset decision priorities are, from high to low, regular model, entity matching model, and deep learning classification model.

8. The health consultation diversion method according to claim 2, characterized in that: The determining the target service type of the user based on the preset screening rule and the initial service type includes: Extracting the business scenario of the user; and determining the target service type of the user based on the business scenario and the initial service type.

9. The health consultation diversion method according to any one of claims 2 to 8, characterized in that: After encoding the health consultation complaint information to obtain the user's feature vector, the method further includes: If the first feature vector contains potential health risk information, it is determined that the target service type of the user is a doctor consultation service type.

10. A health consultation diversion device, characterized in that: include: The acquisition module is used to obtain the user's health consultation complaint information; An encoding module, used for encoding the health consultation complaint information to obtain a feature vector of the user; A prediction module, configured to process the feature vector of the user based on a preset prediction model to obtain an initial service type of the user, wherein the preset prediction model includes at least two of an entity matching model, a deep learning classification model, and a regular model; The screening module is used to determine the target service type of the user based on preset screening rules and the initial service type, and send the health consultation complaint information to the task queue corresponding to the target service type.

11. An electronic device, characterized in that: include: processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the health consultation diversion method as described in any one of claims 1-9 by executing the executable instructions.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the health consultation diversion method as described in any one of claims 1-9 is implemented.

Images