Registration recommendation methods, model training methods, devices, electronic equipment and media

Abstract

This application discloses a registration recommendation method, model training method, device, electronic device, and medium, relating to the field of medical service equipment technology. In the registration recommendation model processing, the target primary department can be used as prior knowledge to improve the accuracy of the model. An attention mechanism layer helps the model understand the contextual information of the second feature and selects the feature that contributes most to the registration recommendation, thus improving accuracy. Then, through a fully connected layer and a classifier, the target secondary department is obtained, and both the target primary and secondary departments are displayed to the user. The target secondary department is the department recommended for the user's registration. Thus, this application can automatically generate recommended secondary departments for registration based on the user's voice-based medical information, facilitating registration with high accuracy.

Description

Technical Field

[0001] This application relates to the field of medical service equipment technology, and in particular to a registration recommendation method, a model training method, a device, electronic equipment and a medium. Background Technology

[0002] With the development of information and network technology, various life services that can provide convenience for users have emerged. For example, self-service registration machines in hospitals allow users to register for appointments, thereby improving registration efficiency and avoiding queuing.

[0003] However, comprehensive hospitals have many departments with high specialization, and many departments also have overlapping services. Users who are not familiar with the hospital departments often face the problem of not knowing which department to register with, and self-service registration machines cannot recommend departments to users. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a registration recommendation method, a model training method, a device, electronic equipment, and a medium, which can automatically generate recommended secondary departments for registration based on the user's voice-based medical information, facilitating the user's registration process.

[0005] The registration recommendation method according to the first aspect of this application includes:

[0006] Acquire voice-based medical information and convert it into text-based medical information;

[0007] The medical condition text information is segmented into words to obtain a word set, which includes multiple medical condition words.

[0008] Based on the aforementioned set of terms, the target primary department is determined from multiple primary departments;

[0009] Each disease-related term and the target primary department in the term set are concatenated to obtain the concatenated text.

[0010] The concatenated text is input into a trained registration recommendation model; wherein the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier;

[0011] The first feature is obtained by performing feature extraction processing on the concatenated text through the embedding layer;

[0012] The first feature is subjected to bidirectional sequence feature extraction and context fusion processing through the bidirectional gated recurrent layer to obtain the second feature;

[0013] The second feature is input into the attention mechanism layer to obtain the attention feature;

[0014] The attention feature is concatenated with the second feature to obtain the concatenated feature;

[0015] Through the fully connected layer, the splicing features are associated with each secondary department to obtain multiple associated features;

[0016] The multiple associated features are input into the classifier to obtain the probability distribution of the associated features;

[0017] Based on the probability distribution, the secondary department corresponding to the associated feature with the highest probability value is selected as the target secondary department;

[0018] The target primary department and the target secondary department are displayed.

[0019] The appointment recommendation method according to the embodiments of this application has at least the following beneficial effects: The appointment recommendation method first converts the voice-based medical information into medical text information, performs word segmentation on the medical text information to obtain a word set, determines the target primary department based on the word set, and then concatenates the target primary department with the word set to obtain concatenated text. The concatenated text is then input into a trained appointment recommendation model to obtain the target secondary department. During the processing of the appointment recommendation model, the target primary department can serve as prior knowledge to improve the accuracy of the appointment recommendation model. The embedding layer is used to extract features from the concatenated text to capture semantic information; the bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature, not only extracting local features of the first feature but also fusing global context information through bidirectional processing, thereby enhancing the appointment recommendation model's ability to understand and represent data. The attention mechanism layer helps the appointment recommendation model understand the context information of the second feature and helps the appointment recommendation model select the feature that contributes most to the appointment recommendation, which is beneficial to improving the accuracy of the appointment recommendation model. Then, through the fully connected layer and classifier, the target secondary department is obtained, and the target primary department and target secondary department are displayed to the user. The target secondary department is the department recommended for the user to register. In this way, this application can automatically generate recommended secondary departments for the user based on the user's voice medical information, making it convenient for the user to register, and the accuracy of the recommended departments is high.

[0020] According to some embodiments of the first aspect of this application, the attention mechanism layer includes a channel attention module and a spatial attention module;

[0021] The step of inputting the second feature into the attention mechanism layer to obtain the attention feature includes:

[0022] The second feature is processed by the channel attention module to obtain the channel attention coefficients;

[0023] Multiply the second feature by the channel attention coefficient to obtain the channel attention feature;

[0024] The spatial attention module performs spatial attention processing on the channel attention features to obtain spatial attention coefficients.

[0025] The attention features are obtained by multiplying the channel attention features by the spatial attention coefficients.

[0026] According to some embodiments of the first aspect of this application, determining the target primary department from multiple primary departments based on the set of terms includes:

[0027] The disease-related terms in the term set are matched with multiple preset primary department keyword corpora to obtain multiple matching degree values; wherein, each primary department keyword corpus corresponds one-to-one with a primary department;

[0028] The corpus of keywords corresponding to the primary department with the largest matching degree among multiple primary departments is used as the target keyword corpus. Based on the target keyword corpus, the target primary department is determined from multiple primary departments.

[0029] According to some embodiments of the first aspect of this application, before displaying the target primary department and the target secondary department, the method further includes:

[0030] The target secondary department and the target primary department were tested;

[0031] If it is detected that the target secondary department does not belong to the lower category of the target primary department, the medical condition text information is re-segmented to obtain a new word set, and then the process jumps to the step of determining the target primary department from multiple primary departments based on the word set.

[0032] According to some embodiments of the first aspect of this application, after displaying the target primary department and the target secondary department, the method further includes:

[0033] Displays the "Agree to Registration Department" button;

[0034] In response to the first operation command for the "agree" button of the registration department, the identity information text box is displayed;

[0035] In response to a second operation instruction for the identity information text box, the input text in the identity information text box is obtained;

[0036] The input text, the target primary department, and the target secondary department are sent to the registration system used for registration.

[0037] According to some embodiments of the first aspect of this application, the registration recommendation model is trained through the following steps:

[0038] Obtain the training concatenated text and the corresponding real labels of the secondary departments;

[0039] The trained concatenated text is then input into the initial registration recommendation model;

[0040] The first training feature is obtained by performing feature extraction processing on the training concatenated text through the embedding layer;

[0041] The first training feature is subjected to bidirectional sequence feature extraction and context fusion processing through the bidirectional gated recurrent layer to obtain the second training feature;

[0042] The second training feature is input into the attention mechanism layer to obtain the attention training feature;

[0043] The attention training feature is concatenated with the second training feature to obtain the concatenated training feature;

[0044] Through the fully connected layer, the spliced ​​training features are associated with each of the secondary departments to obtain multiple associated training features;

[0045] The multiple associated training features are input into the classifier to obtain the training probability distribution of the associated training features;

[0046] Based on the training probability distribution, the secondary department corresponding to the associated training feature with the highest probability value is taken as the target training secondary department;

[0047] The loss value is calculated based on the target training secondary department and the real label of the secondary department. The initial registration recommendation model is iteratively updated based on the loss value to obtain the trained registration recommendation model.

[0048] A second aspect of this application provides a method for training a registration recommendation model, comprising:

[0049] Obtain the training concatenated text and the corresponding real labels of the secondary departments;

[0050] The trained concatenated text is input into the initial registration recommendation model; wherein, the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier;

[0051] The first training feature is obtained by performing feature extraction processing on the training concatenated text through the embedding layer;

[0052] The first training feature is subjected to bidirectional sequence feature extraction and context fusion processing through the bidirectional gated recurrent layer to obtain the second training feature;

[0053] The second training feature is input into the attention mechanism layer to obtain the attention training feature;

[0054] The attention training feature is concatenated with the second training feature to obtain the concatenated training feature;

[0055] Through the fully connected layer, the spliced ​​training features are associated with each secondary department to obtain multiple associated training features;

[0056] The multiple associated training features are input into the classifier to obtain the training probability distribution of the associated training features;

[0057] Based on the training probability distribution, the secondary department corresponding to the associated training feature with the highest probability value is taken as the target training secondary department;

[0058] The loss value is calculated based on the target training secondary department and the real label of the secondary department. The initial registration recommendation model is iteratively updated based on the loss value to obtain the trained registration recommendation model.

[0059] A third aspect of this application provides a registration recommendation device, comprising:

[0060] An acquisition unit is used to acquire voice-based medical information and convert the voice-based medical information into text-based medical information.

[0061] The word segmentation unit is used to segment the medical condition text information into words to obtain a word set, which includes multiple medical condition words.

[0062] The determining unit is used to determine the target primary department from multiple primary departments based on the set of words;

[0063] The word splicing unit is used to splice each of the disease-related words and the target primary department in the word set to obtain spliced ​​text;

[0064] An input unit is used to input the concatenated text into a trained registration recommendation model; wherein, the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier;

[0065] An extraction unit is used to perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature;

[0066] The fusion unit is used to perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain the second feature;

[0067] An attention unit is used to input the second feature into the attention mechanism layer to obtain an attention feature;

[0068] The feature splicing unit is used to splice the attention feature and the second feature to obtain the spliced ​​feature;

[0069] The association unit is used to associate the splicing features with each secondary department through the fully connected layer to obtain multiple association features;

[0070] A classification unit is used to input the multiple associated features into the classifier to obtain the probability distribution of the associated features;

[0071] The selection unit is used to select the secondary department corresponding to the associated feature with the highest probability value as the target secondary department based on the probability distribution.

[0072] The display unit is used to display the target primary department and the target secondary department.

[0073] A fourth aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the registration recommendation method described in any one of the first aspects of the embodiment.

[0074] A fifth aspect of this application provides a computer-readable storage medium storing a computer program, characterized in that, when executed by a processor, the computer program implements the registration recommendation method described in any one of the first aspects of the embodiment.

[0075] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0076] The present application will be further described below with reference to the accompanying drawings and embodiments, wherein:

[0077] Figure 1 This is a flowchart illustrating the steps of the registration recommendation method according to an embodiment of this application;

[0078] Figure 2 This is a schematic diagram of the registration recommendation model according to an embodiment of this application;

[0079] Figure 3This is a functional block diagram of the registration recommendation device according to an embodiment of this application;

[0080] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0081] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0082] In the description of this application, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0083] In the description of this application, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0084] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0085] In the description of this application, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0086] The registration recommendation method provided in this application relates to the field of artificial intelligence technology. The registration recommendation method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the registration recommendation method, but is not limited to the above forms.

[0087] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0088] The first aspect of this application provides a method for recommending medical appointments. (Refer to...) Figure 1 , Figure 1 This is a flowchart illustrating the steps of the registration recommendation method according to an embodiment of this application. The registration recommendation method according to an embodiment of this application may include, but is not limited to, the following steps:

[0089] Step S110: Obtain voice medical information and convert the voice medical information into text medical information;

[0090] In one embodiment, the registration recommendation method can be applied to a hospital's self-service registration device, which is equipped with a microphone to collect the user's voice and thus obtain voice-based medical information.

[0091] In other embodiments, the appointment recommendation method can be applied to the user's terminal device, such as a mobile phone, laptop, or tablet computer, to collect the user's voice through a microphone on the user's terminal device, thereby obtaining voice-based medical information.

[0092] It is worth noting that after obtaining the voice-based medical information, speech recognition is performed on it to obtain the text-based medical information. For example, the voice-based medical information can be input into the Whisper-Tiny model, which outputs the voice-text medical information. The Whisper-Tiny model is an end-to-end deep learning model with multilingual and multi-task capabilities, and can be used for various speech processing tasks, including speech-to-text, speech translation, and speaker identification.

[0093] Step S120: Perform word segmentation on the medical condition text information to obtain a word set, which includes multiple medical condition words;

[0094] In one embodiment, the jieba segmentation algorithm is used for word segmentation. The principle of jieba segmentation is mainly based on Hidden Markov Models (HMMs) and Trie structures. It first performs preliminary segmentation of the text using a Trie, obtaining a series of possible candidate words. Then, the HMM is used to evaluate the probability of these candidate words and select the optimal segmentation result. This method, combining dictionaries and statistical models, enables jieba segmentation to achieve high accuracy and efficiency. For example, a medical condition text message "I have persistent abdominal pain today" will, after segmentation using the jieba algorithm, yield multiple condition words: "today," "persistent," and "abdominal pain," resulting in a word set {"today," "persistent," "abdominal pain"}.

[0095] Step S130: Based on the word set, determine the target primary department from multiple primary departments;

[0096] In one embodiment, in a hospital, departments are typically divided into multiple primary departments, each of which includes multiple secondary departments. For example, primary departments may include internal medicine, surgery, pediatrics, obstetrics and gynecology, and ENT. Internal medicine may include the following secondary departments: respiratory medicine, gastroenterology, cardiology, neurology, nephrology, and endocrinology. Surgery may include the following secondary departments: general surgery, neurosurgery, cardiothoracic surgery, urology, orthopedics, and plastic surgery. Pediatrics may include the following secondary departments: neonatology and pediatric internal medicine. Obstetrics and gynecology may include the following secondary departments: gynecology and obstetrics. ENT may include the following secondary departments: otolaryngology, ophthalmology, and dentistry.

[0097] Step S140: Combine each disease-related word and the target primary department in the word set to obtain the combined text;

[0098] Step S150: Input the concatenated text into the trained registration recommendation model; wherein, the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier;

[0099] Step S160: The concatenated text is processed by feature extraction through the embedding layer to obtain the first feature;

[0100] Step S170: The first feature is subjected to bidirectional sequence feature extraction and context fusion processing through a bidirectional gated recurrent layer to obtain the second feature;

[0101] Step S180: Input the second feature into the attention mechanism layer to obtain the attention feature;

[0102] Step S190: Concatenate the attention feature and the second feature to obtain the concatenated feature;

[0103] Step S200: Through a fully connected layer, the spliced ​​features are associated with each secondary department to obtain multiple associated features;

[0104] Step S210: Input multiple associated features into the classifier to obtain the probability distribution of the associated features;

[0105] Step S220: Based on the probability distribution, the secondary department corresponding to the associated feature with the highest probability value is selected as the target secondary department;

[0106] It is worth noting that, referring to Figure 2 , Figure 2 This is a schematic diagram of the registration recommendation model according to an embodiment of this application. In the registration recommendation model, ALBERT is used as the embedding layer. ALBERT (A Lite BERT) is a lightweight variant of BERT used to generate embedded representations of the input data. The embedding layer performs feature extraction on the concatenated text to obtain a fixed-size embedding vector, which is the first feature. The first feature retains the original information of the concatenated text while incorporating semantic context.

[0107] In step S170 of some embodiments, the Bidirectional Gated Recurrent Unit (BRU) utilizes two independent GRU networks to process the forward and backward traversals of the input sequence (first feature) respectively, thereby enabling a more comprehensive understanding of the order of symptom occurrence and their interrelationships. The BRU not only extracts local features from the first feature (i.e., features at each time step) but also integrates global contextual information through bidirectional processing, thus enhancing the registration recommendation model's ability to understand and represent data.

[0108] In step S200 of some embodiments, the process by which the fully connected layer associates the splicing features with each secondary department can be represented as follows:

[0109] ;

[0110] in, Representation updates feature vectors, The `fully_connected` property represents the association feature between the spliced ​​feature and the i-th secondary department. `fully_connected` is a function that represents the fully connected mapping and is used to map and associate the spliced ​​feature with the secondary department.

[0111] For example, the set of secondary departments is [a1, a2, ..., an], which includes all secondary departments, where a1 to an represent different secondary departments. The concatenation feature is X, and the function of the fully connected mapping can be the concatenation function. Then, in the fully connected layer, Let [X, ai], where i is a positive integer greater than or equal to 1 and less than or equal to n, and n is the total number of secondary departments.

[0112] In step S210 of some embodiments, a Softmax function is used as a classifier. The Softmax function can transform the original output into a probability distribution such that the output value of each class is between 0 and 1, and the sum of the probabilities of all classes is 1.

[0113] Step S230: Display the target primary department and the target secondary department.

[0114] The appointment recommendation method in this embodiment, through steps S110 to S230, first converts the voice-based medical information into text-based medical information, performs word segmentation on the text-based medical information to obtain a word set, determines the target primary department based on the word set, and then concatenates the target primary department with the word set to obtain concatenated text. This concatenated text is then input into a trained appointment recommendation model to obtain the target secondary department. During the processing of the appointment recommendation model, the target primary department can serve as prior knowledge to improve the accuracy of the model. The embedding layer is used to extract features from the concatenated text to capture semantic information. The bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature, not only extracting local features of the first feature but also fusing global contextual information through bidirectional processing, thereby enhancing the appointment recommendation model's ability to understand and represent data. The attention mechanism layer helps the appointment recommendation model understand the contextual information of the second feature and helps it select the feature that contributes most to the appointment recommendation, which is beneficial to improving the accuracy of the model. Then, through the fully connected layer and classifier, the target secondary department is obtained, and the target primary department and target secondary department are displayed to the user. The target secondary department is the department recommended for the user to register. In this way, this application can automatically generate recommended secondary departments for the user based on the user's voice medical information, making it convenient for the user to register, and the accuracy of the recommended departments is high.

[0115] In one embodiment, reference is made to Figure 2 The attention mechanism layer includes a channel attention module and a spatial attention module. Step S180 may include, but is not limited to, steps S310 to S340.

[0116] Step S310: Perform channel attention processing on the second feature through the channel attention module to obtain the channel attention coefficients;

[0117] Step S320: Multiply the second feature by the channel attention coefficient to obtain the channel attention feature;

[0118] Step S330: Spatial attention processing is performed on the channel attention features through the spatial attention module to obtain spatial attention coefficients;

[0119] Step S340: Multiply the channel attention features by the spatial attention coefficients to obtain the attention features.

[0120] It is worth noting that steps S310 to S340 implement channel attention and spatial attention processing for the second feature. The channel attention module helps the model select the features that contribute most to the registration recommendation. For example, if certain features (such as age, gender, specific symptoms, etc.) are highly correlated with the registration department, the channel attention mechanism can assign higher weights to these features; the spatial attention module helps the model focus on the parts of the text that describe key information such as symptoms and medical history.

[0121] Specifically, the processing procedure of the channel attention module can be represented as follows:

[0122] ;

[0123] Where Mc(Y1) represents the channel attention coefficient obtained by performing channel attention processing on the second feature; Y1 represents the first activation feature vector; and Y2 represents the channel attention feature.

[0124] The processing procedure of the spatial attention module can be represented as follows:

[0125] ;

[0126] Where Ms(Y2) represents the spatial attention coefficient obtained by performing spatial attention processing on the channel attention features; Y2 represents the channel attention features; and Y3 represents the attention features.

[0127] In some embodiments, step S130 may include, but is not limited to, steps S410 to S420.

[0128] Step S410: Match the disease-related words in the word set with multiple preset primary department keyword corpora to obtain multiple matching degree values; wherein, the primary department keyword corpora correspond one-to-one with the primary departments;

[0129] Specifically, the primary department key corpus includes multiple medical record terms, which are obtained from the medical records of patients in the corresponding primary department. For example, a primary department key corpus might be a surgical key corpus, which includes multiple surgical-related case terms. Medical record terms are collected from the medical records of multiple surgical patients, for example, by collecting high-frequency words from multiple medical records, thus generating the surgical key corpus. It should be noted that patient permission must be obtained before collecting medical record terms from patients' medical records.

[0130] In another embodiment, the keyword corpus for primary departments is formed by multiple doctors from the corresponding primary departments, and this application does not limit this.

[0131] It is worth noting that each condition term in the word set is matched with the key corpus of each primary department. If a word with the same condition term appears in the key corpus of the primary department, it means that the condition term has been successfully matched with the key corpus of the primary department. The number of condition terms in the word set that have been successfully matched with the key corpus of the primary department is taken as the matching degree value between the word set and the key corpus of the primary department.

[0132] For example, a primary departmental keyword corpus might be an ENT keyword corpus, with the word set {"sore throat", "headache", "nasal congestion", "sinusitis"}. If every word in this set matches the ENT keyword corpus, then the match degree between this word set and the ENT keyword corpus is 4. As another example, if the word set is {"sore throat", "headache", "nasal congestion", "sinusitis"}, and the primary departmental keyword corpus is an internal medicine keyword corpus, and only "headache" matches the internal medicine keyword corpus, then the match degree between this word set and the internal medicine keyword corpus is 1.

[0133] Step S420: The corpus of primary department keywords corresponding to the largest matching value among multiple primary departments is used as the target keyword corpus. Based on the target keyword corpus, the target primary department is determined from multiple primary departments.

[0134] For example, the matching degree of the keyword corpus for ENT is 4, while the matching degree of the keyword corpus for other first-level departments is also 4. Therefore, the keyword corpus for ENT is used as the target keyword corpus, and the first-level department corresponding to the keyword corpus for ENT is ENT. Therefore, ENT is used as the target first-level department.

[0135] In one embodiment, the registration recommendation method of this application includes steps S510 to S520 before step S230.

[0136] Step S510: Conduct testing on the target secondary departments and the target primary departments;

[0137] Step S520: If the target secondary department is detected to be a sub-category of the target primary department, the medical condition text information is re-segmented to obtain a new word set, and the process jumps to determine the target primary department from multiple primary departments based on the word set.

[0138] Specifically, when a target secondary department is detected as not belonging to a subcategory of a target primary department (e.g., the target primary department is surgery, but the target secondary department is neonatology, which is not a subcategory of surgery), this may be due to an error in determining the target primary department in step S130 or an error in model prediction. Therefore, the medical text information is re-segmented to obtain a new word set. Based on the new word set, steps S130 to S220 are repeated to obtain new target secondary and primary departments. This ensures the accuracy of appointment recommendations. When a target secondary department is detected as belonging to a subcategory of a target primary department, step S230 can be executed directly.

[0139] In one embodiment, the registration recommendation method of this application includes steps S610 to S640 after step S230.

[0140] Step S610: Display the "Agree to Registration Department" button;

[0141] Step S620: In response to the first operation instruction for the registration department consent button, display the identity information text box;

[0142] Step S630: In response to the second operation instruction for the identity information text box, obtain the input text in the identity information text box;

[0143] Step S640: Send the input text, target primary department, and target secondary department to the registration system used for registration.

[0144] In one embodiment, the method of this application is applied to a self-service registration device. The self-service registration device is equipped with a touch screen displaying the target secondary department and the target primary department, as well as a "Register Department Agree" button. When the user presses the "Register Department Agree" button on the touch screen, a first operation instruction is generated for the "Register Department Agree" button, and then an identity information text box is displayed on the touch screen. The user then enters their identity information in the identity information text box, thereby generating a second operation instruction for the identity information text box. The entered text, the target primary department, and the target secondary department are sent to the registration system used for registration to help the user complete the registration and improve convenience. It should be noted that this application does not specifically limit the registration system; an existing registration system can be directly used, and the user's identity information is obtained only after obtaining the user's permission.

[0145] In one embodiment, the training steps of the registration recommendation model include, but are not limited to, steps S710 to S800.

[0146] Step S710: Obtain the training spliced ​​text and the corresponding real labels of the secondary departments;

[0147] It should be noted that the training spliced ​​text can consist of preset words used to characterize the disease and preset primary departments. Those skilled in the art can construct the training spliced ​​text according to actual needs.

[0148] Step S720: Input the trained concatenated text into the initial registration recommendation model;

[0149] Step S730: The training spliced ​​text is processed by feature extraction through the embedding layer to obtain the first training features;

[0150] Step S740: The first training feature is subjected to bidirectional sequence feature extraction and context fusion processing through a bidirectional gated recurrent layer to obtain the second training feature;

[0151] Step S750: Input the second training feature into the attention mechanism layer to obtain the attention training feature;

[0152] In one embodiment, step S750 specifically includes the following steps:

[0153] The second training feature is processed by channel attention module to obtain channel attention training coefficients.

[0154] Multiply the second training feature by the channel attention training coefficient to obtain the channel attention training feature;

[0155] Spatial attention training coefficients are obtained by performing spatial attention processing on the channel attention training features using a spatial attention module.

[0156] The attention training features are obtained by multiplying the channel attention training features by the spatial attention training coefficients.

[0157] Step S760: The attention training features and the second training features are concatenated to obtain the concatenated training features;

[0158] Step S770: Through a fully connected layer, the spliced ​​training features are associated with each secondary department to obtain multiple associated training features;

[0159] Step S780: Input multiple associated training features into the classifier to obtain the training probability distribution of the associated training features;

[0160] Step S790: Based on the training probability distribution, the secondary department corresponding to the associated training feature with the highest probability value is taken as the target training secondary department;

[0161] It should be noted that the processing procedure for training concatenated text in the registration recommendation model is the same as the processing procedure for training concatenated text in the registration recommendation model.

[0162] Step S800: Calculate the loss value based on the target training secondary departments and the real labels of the secondary departments, and iteratively update the initial registration recommendation model based on the loss value to obtain the trained registration recommendation model.

[0163] It should be noted that after obtaining the target training secondary department, a preset loss function is used to calculate the loss value. After obtaining the loss value, the initial registration recommendation model is updated based on the loss value. This process is repeated until the preset number of iterations is reached, thereby obtaining the trained registration recommendation model. In one embodiment, the cross-entropy loss function is used as the loss function. The cross-entropy loss function is typically used for classification tasks and is suitable for scenarios where the target variable is discrete (such as the real label of the secondary department in this application).

[0164] A second aspect of this application provides a method for training a registration recommendation model, comprising the following steps:

[0165] Obtain the training concatenated text and the corresponding real labels for the secondary departments;

[0166] The training concatenated text is input into the initial registration recommendation model; the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier;

[0167] The first training feature is obtained by performing feature extraction on the training concatenated text through the embedding layer.

[0168] The second training feature is obtained by performing bidirectional sequence feature extraction and context fusion processing on the first training feature through a bidirectional gated recurrent layer.

[0169] The second training feature is input into the attention mechanism layer to obtain the attention training feature;

[0170] The attention training features and the second training features are concatenated to obtain the concatenated training features.

[0171] By using a fully connected layer, the spliced ​​training features are associated with each secondary department to obtain multiple associated training features;

[0172] Multiple associated training features are input into the classifier to obtain the training probability distribution of the associated training features;

[0173] Based on the training probability distribution, the secondary department corresponding to the associated training feature with the highest probability value is used as the target training secondary department;

[0174] The loss value is calculated based on the target training of secondary departments and the real labels of secondary departments. The initial registration recommendation model is iteratively updated based on the loss value to obtain a trained registration recommendation model.

[0175] A third aspect of this application provides a registration recommendation device. (Refer to...) Figure 3 , Figure 3 This is a functional block diagram of the registration recommendation device according to an embodiment of this application. The registration recommendation device includes:

[0176] The acquisition unit 310 is used to acquire voice-based medical information and convert the voice-based medical information into text-based medical information.

[0177] Word segmentation unit 320 is used to segment the medical condition text information into words to obtain a word set, which includes multiple medical condition words;

[0178] Unit 330 is used to determine the target primary department from multiple primary departments based on a set of words;

[0179] Word splicing unit 340 is used to splice each disease word and target primary department in the word set to obtain spliced ​​text;

[0180] Input unit 350 is used to input the concatenated text into the trained registration recommendation model; wherein, the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer and a classifier;

[0181] Extraction unit 360 is used to perform feature extraction processing on the concatenated text through the embedding layer to obtain the first feature;

[0182] Fusion unit 370 is used to perform bidirectional sequence feature extraction and context fusion processing on the first feature through a bidirectional gated recurrent layer to obtain the second feature;

[0183] Attention unit 380 is used to input the second feature into the attention mechanism layer to obtain attention features;

[0184] Feature splicing unit 390 is used to splice the attention feature and the second feature to obtain spliced ​​features;

[0185] The association unit 400 is used to associate the spliced ​​features with each secondary department through a fully connected layer to obtain multiple associated features;

[0186] The classification unit 410 is used to input multiple associated features into the classifier to obtain the probability distribution of the associated features;

[0187] Selection unit 420 is used to select the secondary department corresponding to the associated feature with the highest probability value as the target secondary department based on the probability distribution;

[0188] Display unit 430 is used to display the target primary department and the target secondary department.

[0189] The appointment recommendation device of this application embodiment is used to execute the appointment recommendation method of the first aspect embodiment. When executing the method, the voice-based medical information is first converted into medical text information. The medical text information is then segmented to obtain a word set. Based on the word set, a target primary department is determined. Then, the target primary department is concatenated with the word set to obtain concatenated text. The concatenated text is input into a trained appointment recommendation model to obtain a target secondary department. During the processing of the appointment recommendation model, the target primary department can serve as prior knowledge to improve the accuracy of the appointment recommendation model. The embedding layer is used to extract features from the concatenated text to capture semantic information. The bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature. This not only extracts the local features of the first feature but also fuses global context information through bidirectional processing, thereby enhancing the appointment recommendation model's ability to understand and represent data. The attention mechanism layer helps the appointment recommendation model understand the context information of the second feature and helps the model select the feature that contributes most to the appointment recommendation, which is beneficial to improving the accuracy of the appointment recommendation model. Then, through the fully connected layer and classifier, the target secondary department is obtained, and the target primary department and target secondary department are displayed to the user. The target secondary department is the department recommended for the user to register. In this way, this application can automatically generate recommended secondary departments for the user based on the user's voice medical information, making it convenient for the user to register, and the accuracy of the recommended departments is high.

[0190] It should be noted that the specific implementation of this registration recommendation device is basically the same as the specific embodiment of the registration recommendation method described above, and will not be repeated here. Subject to meeting the requirements of the embodiments of this application, the registration recommendation device may also be equipped with other functional units to implement the registration recommendation method in the above embodiments.

[0191] A fourth aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the registration recommendation method of any one of the first aspects of the embodiment. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0192] Reference Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device according to one embodiment. The electronic device includes:

[0193] The processor 401 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0194] The memory 402 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 402 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 402 and is called and executed by the processor 401 using the registration recommendation method of the embodiments of this application.

[0195] Input / output interface 403 is used to implement information input and output;

[0196] The communication interface 404 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0197] Bus 405 transmits information between various components of the device (e.g., processor 401, memory 402, input / output interface 403, and communication interface 404);

[0198] The processor 401, memory 402, input / output interface 403 and communication interface 404 are connected to each other within the device via bus 405.

[0199] According to a fifth aspect of this application, a computer-readable storage medium stores a computer program that, when executed by a processor, implements the registration recommendation method of any one of the first aspects of this application.

[0200] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0201] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0202] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0203] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0205] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0206] It should be understood that in this application, "at least one (item)" means one or more, and "more than one" means two or more. "And / or" is used to describe the mapping relationship between the mapped objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following mapped objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0207] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0208] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0209] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0210] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0211] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for recommending medical appointments, characterized in that, include: Acquire voice-based medical information and convert it into text-based medical information; The medical condition text information is segmented into words to obtain a word set, which includes multiple medical condition words. Based on the aforementioned set of terms, the target primary department is determined from multiple primary departments; Each disease-related term and the target primary department in the term set are concatenated to obtain the concatenated text. The concatenated text is input into a trained registration recommendation model; wherein the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier; The first feature is obtained by performing feature extraction processing on the concatenated text through the embedding layer; The first feature is subjected to bidirectional sequence feature extraction and context fusion processing through the bidirectional gated recurrent layer to obtain the second feature; The second feature is input into the attention mechanism layer to obtain the attention feature; The attention feature is concatenated with the second feature to obtain the concatenated feature; Through the fully connected layer, the splicing features are associated with each secondary department to obtain multiple associated features; The multiple associated features are input into the classifier to obtain the probability distribution of the associated features; Based on the probability distribution, the secondary department corresponding to the associated feature with the highest probability value is selected as the target secondary department; Display the target primary department and the target secondary department; The attention mechanism layer includes a channel attention module and a spatial attention module; The step of inputting the second feature into the attention mechanism layer to obtain the attention feature includes: The second feature is processed by the channel attention module to obtain the channel attention coefficients; Multiply the second feature by the channel attention coefficient to obtain the channel attention feature; The spatial attention module performs spatial attention processing on the channel attention features to obtain spatial attention coefficients. The attention features are obtained by multiplying the channel attention features by the spatial attention coefficients. Before displaying the target primary department and the target secondary department, the method further includes: The target secondary department and the target primary department were tested; If it is detected that the target secondary department does not belong to the lower category of the target primary department, the medical condition text information is re-segmented to obtain a new word set, and then the process jumps to the step of determining the target primary department from multiple primary departments based on the word set.

2. The registration recommendation method according to claim 1, characterized in that, The process of determining the target primary department from multiple primary departments based on the aforementioned word set includes: The disease-related terms in the term set are matched with multiple preset primary department keyword corpora to obtain multiple matching degree values; wherein, each primary department keyword corpus corresponds one-to-one with a primary department; The corpus of keywords corresponding to the primary department with the largest matching degree among multiple primary departments is used as the target keyword corpus. Based on the target keyword corpus, the target primary department is determined from multiple primary departments.

3. The registration recommendation method according to claim 1, characterized in that, After displaying the target primary department and the target secondary department, the method further includes: Displays the "Agree to Registration Department" button; In response to the first operation command for the "agree" button of the registration department, the identity information text box is displayed; In response to a second operation instruction for the identity information text box, the input text in the identity information text box is obtained; The input text, the target primary department, and the target secondary department are sent to the registration system used for registration.

4. The registration recommendation method according to claim 1, characterized in that, The registration recommendation model is trained through the following steps: Obtain the training concatenated text and the corresponding real labels of the secondary departments; The trained concatenated text is then input into the initial registration recommendation model; The first training feature is obtained by performing feature extraction processing on the training concatenated text through the embedding layer; The first training feature is subjected to bidirectional sequence feature extraction and context fusion processing through the bidirectional gated recurrent layer to obtain the second training feature; The second training feature is input into the attention mechanism layer to obtain the attention training feature; The attention training feature is concatenated with the second training feature to obtain the concatenated training feature; Through the fully connected layer, the spliced ​​training features are associated with each of the secondary departments to obtain multiple associated training features; The multiple associated training features are input into the classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, the secondary department corresponding to the associated training feature with the highest probability value is taken as the target training secondary department; The loss value is calculated based on the target training secondary department and the real label of the secondary department. The initial registration recommendation model is iteratively updated based on the loss value to obtain the trained registration recommendation model.

5. A registration recommendation device, characterized in that, include: An acquisition unit is used to acquire voice-based medical information and convert the voice-based medical information into text-based medical information. The word segmentation unit is used to segment the medical condition text information into words to obtain a word set, which includes multiple medical condition words. The determining unit is used to determine the target primary department from multiple primary departments based on the set of words; The word splicing unit is used to splice each of the disease-related words and the target primary department in the word set to obtain spliced ​​text; An input unit is used to input the concatenated text into a trained registration recommendation model; wherein, the registration recommendation model includes an embedding layer, a bidirectional gated recurrent layer, an attention mechanism layer, a fully connected layer, and a classifier; An extraction unit is used to perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature; The fusion unit is used to perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain the second feature; An attention unit is used to input the second feature into the attention mechanism layer to obtain an attention feature; The feature splicing unit is used to splice the attention feature and the second feature to obtain the spliced ​​feature; The association unit is used to associate the splicing features with each secondary department through the fully connected layer to obtain multiple association features; A classification unit is used to input the multiple associated features into the classifier to obtain the probability distribution of the associated features; The selection unit is used to select the secondary department corresponding to the associated feature with the highest probability value as the target secondary department based on the probability distribution. The display unit is used to display the target primary department and the target secondary department; The attention mechanism layer includes a channel attention module and a spatial attention module; The step of inputting the second feature into the attention mechanism layer to obtain the attention feature includes: The second feature is processed by the channel attention module to obtain the channel attention coefficients; Multiply the second feature by the channel attention coefficient to obtain the channel attention feature; The spatial attention module performs spatial attention processing on the channel attention features to obtain spatial attention coefficients. The attention features are obtained by multiplying the channel attention features by the spatial attention coefficients. Before displaying the target primary department and the target secondary department, the method further includes: The target secondary department and the target primary department were tested; If it is detected that the target secondary department does not belong to the lower category of the target primary department, the medical condition text information is re-segmented to obtain a new word set, and then the process jumps to the step of determining the target primary department from multiple primary departments based on the word set.

6. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the registration recommendation method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the registration recommendation method according to any one of claims 1 to 4.

Images