Registration recommendation method and device, model training method and device, electronic equipment and medium

Abstract

The invention discloses a registration recommendation method and device, a model training method and device, electronic equipment and a medium, and relates to the technical field of medical service equipment. According to the method, in the processing process of the registration recommendation model, the target first-level department can serve as priori knowledge, so that the accuracy of the registration recommendation model is improved. And the attention mechanism layer can help the registration recommendation model to understand the context information of the second feature and help the registration recommendation model to select the feature which is most contributive to registration recommendation, so that the accuracy of the registration recommendation model is improved. And then a target second-level department is obtained through the effect of the full connection layer and the classifier, the target first-level department and the target second-level department are displayed to the user, and the target second-level department is the department recommended to the user for registration. Thus, according to the voice disease condition information of the user, the second-level department recommended for registration can be automatically generated for the user, registration of the user is facilitated, and the accuracy of the department recommended for registration is high.

Description

Technical Field

[0001] This application relates to the technical field of medical service equipment, and particularly relates to a registration recommendation method, a model training method, a device, an electronic device, and a medium. Background Art

[0002] With the development of information and network technologies, various life services that can provide convenience for users have emerged. For example, self-service registration devices in hospitals allow users to register through these devices, which can improve the registration efficiency and avoid users queuing.

[0003] However, general hospitals have numerous departments with strong professionalism, and there are also overlapping services in many departments. For users who are not familiar with the hospital departments, they often face the problem of not knowing which department to register for, and self-service registration devices cannot recommend departments to users. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art. For this purpose, this application proposes a registration recommendation method, a model training method, a device, an electronic device, and a medium, which can automatically generate a recommended secondary department for registration based on the user's voice disease condition information, facilitating the user's registration.

[0005] According to the registration recommendation method of the first aspect embodiment of this application, it includes: Obtain voice disease condition information and convert the voice disease condition information into disease condition text information; Perform word segmentation processing on the disease condition text information to obtain a word set, where the word set includes multiple disease condition words; Based on the word set, determine a target primary department from multiple primary departments; Concatenate each disease condition word in the word set and the target primary department to obtain a concatenated text; 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; Perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature; Perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature; Input the second feature into the attention mechanism layer to obtain an attention feature; Concatenate the attention feature and the second feature to obtain a concatenated feature; Through the fully connected layer, associate the concatenated feature with each secondary department respectively to obtain multiple associated features; Input the multiple associated features into the classifier to obtain the probability distribution of the associated features; Based on the probability distribution, use the secondary department corresponding to the associated feature with the largest probability value as the target secondary department; Display the target primary department and the target secondary department.

[0006] According to the registration recommendation method of the embodiments of the present application, it has at least the following beneficial effects: The registration recommendation method first converts the voice medical condition information into medical condition text information, performs word segmentation processing on the medical condition text information to obtain a word set, determines the target primary department based on the word set, then splices the target primary department and the word set to obtain a spliced text, and inputs the spliced text into the trained registration recommendation model to obtain the target secondary department. During the processing of the registration recommendation model, the target primary department can be used as prior knowledge to improve the accuracy of the registration recommendation model. The embedding layer is used to extract features from the spliced text to capture the semantic information in the spliced text; the bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature, not only extracting the local features of the first feature, but also fusing the global context information through bidirectional processing, thereby enhancing the registration recommendation model's ability to understand and represent data. The attention mechanism layer can help the registration recommendation model understand the context information of the second feature and help the registration recommendation model select the features that contribute the most to the registration recommendation, which is beneficial to improving the accuracy of the registration recommendation model. Then, through the action of the fully connected layer and the classifier, the target secondary department is obtained, and the target primary department and the target secondary department are displayed to the user. The target secondary department is the department recommended for the user to register. In this way, the present application can automatically generate the secondary department recommended for the user to register according to the user's voice medical condition information, which is convenient for the user to register, and the accuracy of the recommended registration department is high.

[0007] According to some embodiments of the first aspect of the present application, 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: Perform channel attention processing on the second feature through the channel attention module to obtain a channel attention coefficient; Multiply the second feature by the channel attention coefficient to obtain a channel attention feature; Perform spatial attention processing on the channel attention feature through the spatial attention module to obtain a spatial attention coefficient; Multiply the channel attention feature by the spatial attention coefficient to obtain the attention feature.

[0008] According to some embodiments of the first aspect of the present application, determining the target first-level department from multiple first-level departments based on the word set includes: Matching the disease condition words in the word set with multiple preset first-level department keyword corpora to obtain multiple matching degree values; wherein, the first-level department keyword corpora correspond to the first-level departments one by one; Taking the first-level department keyword corpus corresponding to the largest value among the multiple matching degree values as the target keyword corpus, and determining the target first-level department from the multiple first-level departments based on the target keyword corpus.

[0009] According to some embodiments of the first aspect of the present application, before displaying the target first-level department and the target second-level department, it further includes: Detecting the target second-level department and the target first-level department; In the case where it is detected that the target second-level department does not belong to the subordinate classification of the target first-level department, re-segmenting the disease condition text information to obtain a new word set, and jumping to determining the target first-level department from multiple first-level departments based on the word set.

[0010] According to some embodiments of the first aspect of the present application, after displaying the target first-level department and the target second-level department, it further includes: Displaying a registration department approval button; In response to a first operation instruction for the registration department approval button, displaying an identity information text box; In response to a second operation instruction for the identity information text box, obtaining the input text in the identity information text box; Sending the input text, the target first-level department, and the target second-level department to a registration system for registration.

[0011] According to some embodiments of the first aspect of the present application, the registration recommendation model is trained through the following steps: Obtaining a training spliced text and a true label of the second-level department corresponding to the training spliced text; Inputting the training spliced text into an initial registration recommendation model; Performing feature extraction processing on the training spliced text through the embedding layer to obtain a first training feature; Performing bidirectional sequence feature extraction and context fusion processing on the first training feature through the bidirectional gated recurrent layer to obtain a second training feature; Inputting the second training feature into the attention mechanism layer to obtain an attention training feature; Concatenate the attention training feature and the second training feature to obtain a concatenated training feature; Through the fully connected layer, associate the concatenated training feature with each of the secondary departments respectively to obtain a plurality of associated training features; Input the plurality of associated training features into the classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, use the secondary department corresponding to the associated training feature with the largest probability value as the target training secondary department; Calculate a loss value based on the target training secondary department and the true label of the secondary department, and iteratively update the initial registration recommendation model based on the loss value to obtain the trained registration recommendation model.

[0012] An embodiment of the second aspect of the present application provides a method for training a registration recommendation model, including: Obtain training concatenated text and the true label of the secondary department corresponding to the training concatenated text; Input the training concatenated text into an 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; Extract features from the training concatenated text through the embedding layer to obtain a first training feature; Perform bidirectional sequence feature extraction and context fusion processing on the first training feature through the bidirectional gated recurrent layer to obtain a second training feature; Input the second training feature into the attention mechanism layer to obtain an attention training feature; Concatenate the attention training feature and the second training feature to obtain a concatenated training feature; Through the fully connected layer, associate the concatenated training feature with each secondary department respectively to obtain a plurality of associated training features; Input the plurality of associated training features into the classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, use the secondary department corresponding to the associated training feature with the largest probability value as the target training secondary department; Calculate a loss value based on the target training secondary department and the true label of the secondary department, and iteratively update the initial registration recommendation model based on the loss value to obtain the trained registration recommendation model.

[0013] An embodiment of the third aspect of the present application provides a registration recommendation device, including: An acquisition unit, configured to acquire voice condition information and convert the voice condition information into condition text information; A word segmentation unit, configured to perform word segmentation processing on the condition text information to obtain a word set, where the word set includes a plurality of condition words; A determination unit, configured to determine a target first-level department from a plurality of first-level departments based on the word set; A word splicing unit, configured to splice each of the condition words and the target first-level department in the word set to obtain a spliced text; An input unit, configured to input the spliced 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, configured to perform feature extraction processing on the spliced text through the embedding layer to obtain a first feature; A fusion unit, configured to perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature; An attention unit, configured to input the second feature into the attention mechanism layer to obtain an attention feature; A feature splicing unit, configured to splice the attention feature and the second feature to obtain a spliced feature; An association unit, configured to associate the spliced feature with each second-level department respectively through the fully connected layer to obtain a plurality of associated features; A classification unit, configured to input the plurality of associated features into the classifier to obtain a probability distribution of the associated features; A selection unit, configured to, based on the probability distribution, use the second-level department corresponding to the associated feature with the largest probability value as the target second-level department; A display unit, configured to display the target first-level department and the target second-level department.

[0014] An embodiment of the fourth aspect of the present application provides an electronic device, where the electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the registration recommendation method according to any one of the embodiments of the first aspect is implemented.

[0015] An embodiment of the fifth aspect of the present application provides a computer-readable storage medium, where the storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, the registration recommendation method according to any one of the embodiments of the first aspect is implemented.

[0016] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Description of the Drawings

[0017] The present application will be further described below in conjunction with the drawings and embodiments, where: Figure 1 is a schematic flow chart of the steps of the registration recommendation method according to an embodiment of the present application; Figure 2 is a schematic structural diagram of the registration recommendation model according to an embodiment of the present application; Figure 3 is a functional module block diagram of the registration recommendation device according to an embodiment of the present application; Figure 4 is a schematic structural diagram of the electronic device according to an embodiment of the present application. Detailed Embodiments

[0018] The embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the drawings, where 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 drawings are exemplary only for explaining the present application and should not be construed as limiting the present application.

[0019] In the description of the present application, it should be understood that the orientation or positional relationship indicated by terms such as up, down, front, back, left, right, etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present application.

[0020] In the description of the present application, the meaning of several is more than one, the meaning of multiple is more than two, and understandings such as greater than, less than, exceeding, etc. do not include the present number, and understandings such as above, below, within, etc. include the present number. If there is a description of first and second, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features.

[0021] In the description of the present application, unless otherwise clearly defined, words such as setting, installing, connecting, etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present application in combination with the specific content of the technical solution.

[0022] In the description of this application, the description with reference to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0023] The registration recommendation method provided by the embodiments of this application relates to the field of artificial intelligence technology. The registration recommendation method provided by the embodiments of this application can be applied to a terminal, or to a server side, or can also be software running on a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, or can also be configured as a cloud server providing 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, and big data and artificial intelligence platforms; the software can be an application that implements the registration recommendation method, etc., but is not limited to the above forms.

[0024] This application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. 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 a distributed computing environment, where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0025] The first aspect embodiment of this application provides a registration recommendation method. Refer to Figure 1 , Figure 1 which is the schematic diagram of the step flow of the registration recommendation method of the embodiments of this application. The registration recommendation method of the embodiments of this application can include but is not limited to the following steps: Step S110, obtain voice condition information and convert the voice condition information into condition text information; In one embodiment, the registration recommendation method can be applied to the self-service registration device in a hospital. A microphone is provided on the self-service registration device, and the voice of the user is collected through the microphone, so as to obtain voice condition information.

[0026] In other embodiments, the registration recommendation method can be applied to the user's terminal device, such as electronic devices like mobile phones, laptops, or tablets. The voice of the user is collected through the microphone on the user's terminal device, so as to obtain voice condition information.

[0027] It should be noted that after obtaining the voice condition information, voice recognition is performed on the voice condition information to obtain condition text information. For example, the voice condition information is input into the Whisper-Tiny model, and the Whisper-Tiny model outputs the condition voice text information. The Whisper-Tiny model is an end-to-end deep learning model with multi-language and multi-task capabilities, and can be used for various speech processing tasks, including speech-to-text (transcription), speech translation (translation), and speaker identification.

[0028] Step S120: Perform word segmentation processing on the condition text information to obtain a word set, and the word set includes multiple condition words; In one embodiment, the jieba word segmentation algorithm is used for word segmentation processing. The principle of jieba word segmentation is mainly based on the Hidden Markov Model (HMM) and the Trie structure. It first performs a preliminary segmentation of the text through the Trie to obtain a series of possible vocabulary candidates. Then, the Hidden Markov Model is used to perform probability evaluation on these candidate vocabularies to select the optimal segmentation result. This method combining a dictionary and a statistical model enables jieba word segmentation to reach a relatively high level in terms of accuracy and efficiency. Exemplarily, a condition text information is "I have had continuous abdominal pain today". After performing word segmentation processing using the jieba word segmentation algorithm, multiple condition words are obtained, namely "today", "continuous", and "abdominal pain", and the resulting word set is {"today", "continuous", "abdominal pain"}.

[0029] Step S130: Based on the word set, determine the target first-level department from multiple first-level departments; In one embodiment, in a hospital, departments are usually divided into multiple first-level departments, and each first-level department includes multiple second-level departments. For example, the first-level departments include internal medicine, surgery, pediatrics, obstetrics and gynecology, and otolaryngology. The second-level departments included in internal medicine are respectively respiratory medicine, digestive medicine, cardiovascular medicine, neurology, nephrology, and endocrinology. The second-level departments included in surgery are respectively general surgery, neurosurgery, cardiothoracic surgery, urology, orthopedics, and plastic surgery. The second-level departments included in pediatrics are respectively neonatology and pediatrics. The second-level departments included in obstetrics and gynecology are respectively gynecology and obstetrics. The second-level departments included in otolaryngology are respectively otolaryngology, ophthalmology, and stomatology.

[0030] Step S140: Concatenate each disease condition word in the word set with the target first-level department to obtain a concatenated text. 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. Step S160: Perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature. Step S170: Perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature. Step S180: Input the second feature into the attention mechanism layer to obtain an attention feature. Step S190: Concatenate the attention feature and the second feature to obtain a concatenated feature. Step S200: Through the fully connected layer, associate the concatenated feature with each second-level department respectively to obtain multiple associated features. Step S210: Input the multiple associated features into the classifier to obtain the probability distribution of the associated features. Step S220: Based on the probability distribution, use the second-level department corresponding to the associated feature with the largest probability value as the target second-level department. It should be noted that with reference to Figure 2 , Figure 2 is a schematic structural diagram of the registration recommendation model of the embodiment of the present application. In the registration recommendation model, ALBERT is used as the embedding layer. ALBERT (A Lite BERT) is a lightweight variant of BERT, which is used to generate the embedding representation of the input data. Through the embedding layer, feature extraction processing is performed on the concatenated text to obtain a fixed-size embedding vector, which is the first feature. The first feature not only retains the original information of the concatenated text but also incorporates semantic context.

[0031] In step S170 of some embodiments, the Bidirectional Gated Recurrent Unit processes the forward and reverse traversals of the input sequence (the first feature) using two independent GRU networks respectively, so as to be able to more comprehensively understand the order of appearance of symptoms and their interrelationships. The bidirectional gated recurrent layer not only extracts the local features in the first feature (i.e., the features at each time step), but also fuses the global context information through bidirectional processing, thereby enhancing the data understanding and representation ability of the registration recommendation model.

[0032] In step S200 of some embodiments, the process of the fully connected layer associating the concatenated features with each secondary department can be expressed as: ; where represents the updated feature vector, represents the association feature of the concatenated feature associated with the i-th secondary department, and fully_connected represents the function of the fully connected mapping, which is used to map and associate the concatenated feature with the secondary department.

[0033] Exemplarily, the set of secondary departments is [a1, a2,..., an], which includes all secondary departments, a1 to an respectively represent different secondary departments, the concatenated feature is X, and the function of the fully connected mapping can be the concatenation function. Then, in the fully connected layer, is [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.

[0034] In step S210 of some embodiments, the Softmax function is used as the classifier. The Softmax function can convert the original output into a probability distribution, so that the output value of each category is between 0 and 1, and the sum of the probabilities of all categories is 1.

[0035] Step S230, display the target first-level department and the target second-level department.

[0036] The registration recommendation method according to the embodiments of the present application, through the above steps S110 to S230, first converts the voice condition information into condition text information, performs word segmentation processing on the condition text information to obtain a set of words, determines the target first-level department based on the set of words, then splices the target first-level department and the set of words to obtain a spliced text, inputs the spliced text into the trained registration recommendation model, and obtains the target second-level department. During the processing of the registration recommendation model, the target first-level department can serve as prior knowledge to improve the accuracy of the registration recommendation model. The embedding layer is used to extract features from the spliced text to capture the semantic information in the spliced text; the bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature, not only extracting the local features of the first feature, but also fusing the global context information through bidirectional processing, thereby enhancing the registration recommendation model's ability to understand and represent data. The attention mechanism layer can help the registration recommendation model understand the context information of the second feature and help the registration recommendation model select the features that contribute the most to the registration recommendation, which is beneficial to improving the accuracy of the registration recommendation model. Then, through the action of the fully connected layer and the classifier, the target second-level department is obtained, and the target first-level department and the target second-level department are displayed to the user. The target second-level department is the department recommended for the user to register. In this way, the present application can automatically generate the second-level department recommended for the user to register according to the user's voice condition information, which is convenient for the user to register, and the accuracy of the recommended registration department is high.

[0037] In one embodiment, referring 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.

[0038] Step S310, perform channel attention processing on the second feature through the channel attention module to obtain a channel attention coefficient; Step S320, multiply the second feature by the channel attention coefficient to obtain a channel attention feature; Step S330, perform spatial attention processing on the channel attention feature through the spatial attention module to obtain a spatial attention coefficient; Step S340, multiply the channel attention feature by the spatial attention coefficient to obtain an attention feature.

[0039] It should be noted that through steps S310 to S340, the channel attention processing and spatial attention processing of the second feature are realized. The channel attention module can help the model select the features that contribute the most to the registration recommendation. For example, if certain features (such as age, gender, specific symptoms, etc.) have a higher correlation with the registration department, the channel attention mechanism can assign higher weights to these features; the spatial attention module can help the model focus on the parts of the text that describe key information such as symptoms and medical history.

[0040] Specifically, the processing process of the channel attention module can be expressed as: ; Among them, Mc(Y1) represents the channel attention coefficient obtained by performing channel attention processing on the second feature; Y1 represents the first activation feature vector; Y2 represents the channel attention feature.

[0041] The processing process of the spatial attention module can be expressed as: ; Among them, Ms(Y2) represents the spatial attention coefficient obtained by performing spatial attention processing on the channel attention feature; Y2 represents the channel attention feature; Y3 represents the attention feature.

[0042] In some embodiments, step S130 may include but is not limited to steps S410 to S420.

[0043] Step S410: Match the disease condition words in the word set with multiple preset first-level department keyword corpora to obtain multiple matching degree values; among them, the first-level department keyword corpora correspond to the first-level departments one by one; Specifically, the first-level department keyword corpus includes multiple medical record words, and the medical record words are obtained from the medical record books of patients in the first-level department corresponding to the first-level department keyword corpus. Exemplarily, a first-level department keyword corpus is a surgical keyword corpus, and the surgical keyword corpus includes multiple case words related to surgery. Collect medical record words from the medical record books of multiple surgical patients, such as collecting high-frequency words in multiple medical record books, so as to generate a surgical keyword corpus. It should be noted that when collecting medical record words from the medical record books of patients, the permission of the patients needs to be obtained before proceeding.

[0044] In another embodiment, the first-level department keyword corpus is formed by multiple doctors corresponding to the first-level department, and the present application does not make any limitations on this.

[0045] It should be noted that each disease condition word in the word set is respectively matched with each first-level department keyword corpus. If there is a word in the first-level department keyword corpus that is the same as the disease condition word, it means that the disease condition word is successfully matched with the first-level department keyword corpus, and the number of disease condition words in the word set that are successfully matched with the first-level department keyword corpus is used as the matching degree value between the word set and the first-level department keyword corpus.

[0046] Exemplarily, the key corpus of the first-level department is the key corpus of the otolaryngology department, and the word set is {"sore throat", "headache", "nasal congestion", "sinusitis"}. In this word set, each word can be successfully matched with the key corpus of the otolaryngology department, so the matching degree value of this word set and the key corpus of the otolaryngology department is 4. Another example, the word set is {"sore throat", "headache", "nasal congestion", "sinusitis"}, and the key corpus of the first-level department is the key corpus of the internal medicine department. Only the word "headache" in the word set is successfully matched with the key corpus of the internal medicine department, so the matching degree value of this word set and the key corpus of the internal medicine department is 1.

[0047] Step S420: Use the key corpus of the first-level department corresponding to the largest value among the multiple matching degree values as the target key corpus, and based on the target key corpus, determine the target first-level department from multiple first-level departments.

[0048] For example, the matching degree value corresponding to the key corpus of the otolaryngology department is 4, while the matching degree values corresponding to the key corpora of the other first-level departments are 4. Therefore, the key corpus of the otolaryngology department is used as the target key corpus, and the first-level department corresponding to the key corpus of the otolaryngology department is the otolaryngology department. Therefore, the otolaryngology department is used as the target first-level department.

[0049] In one embodiment, the registration recommendation method of the embodiment of the present application includes steps S510 to S520 before step S230.

[0050] Step S510: Detect the target second-level department and the target first-level department. Step S520: In the case where it is detected that the target second-level department does not belong to the subordinate classification of the target first-level department, re-segment the disease condition text information to obtain a new word set, and jump to determining the target first-level department from multiple first-level departments based on the word set.

[0051] Specifically, when it is detected that the target second-level department does not belong to the subordinate classification of the target first-level department, for example, the target first-level department is the surgical department, and the target second-level department is the neonatal department. The neonatal department is not a subordinate classification of the surgical department. This situation may be caused by an incorrect determination of the target first-level department in step S130, or may be caused by a model prediction error. Therefore, re-segment the disease condition text information to obtain a new word set, and based on the new word set, re-execute steps S130 to S220 to obtain a new target second-level department and a new target first-level department. In this way, the accuracy of the registration recommendation can be ensured. When it is detected that the target second-level department belongs to the subordinate classification of the target first-level department, step S230 can be directly executed.

[0052] In one embodiment, the registration recommendation method of the embodiment of the present application includes steps S610 to S640 after step S230.

[0053] Step S610: Display the approval button for the registration department. Step S620: In response to a first operation instruction for the approval button for the registration department, display an identity information text box. Step S630: In response to a second operation instruction for the identity information text box, obtain the input text in the identity information text box. Step S640: Send the input text, the target first-level department, and the target second-level department to a registration system for registration.

[0054] In one embodiment, the method of the present application is applied to a self-service registration device. The self-service registration device is provided with a touch display screen. The target second-level department and the target first-level department are displayed on the touch display screen, and the approval button for the registration department is also displayed. When the user presses the approval button for the registration department on the touch display screen, a first operation instruction for the approval button for the registration department is generated, and then the identity information text box is displayed on the touch display screen. Then the user inputs the user's identity information in the identity information text box, thereby generating a second operation instruction for the identity information text box. The input text, the target first-level department, and the target second-level department are sent to a registration system for registration, which helps the user complete the registration and improves convenience. It should be noted that the present application does not make specific limitations on the registration system, and an existing registration system can be directly adopted, and the user's identity information is obtained after obtaining the user's permission.

[0055] In one embodiment, the training steps of the registration recommendation model include but are not limited to steps S710 to S800.

[0056] Step S710: Obtain a training spliced text and a true label of the second-level department corresponding to the training spliced text. It should be noted that the training spliced text can be composed of preset words for characterizing the condition and preset first-level departments, and those skilled in the art can construct the training spliced text according to actual needs.

[0057] Step S720: Input the training spliced text into an initial registration recommendation model. Step S730: Perform feature extraction processing on the training spliced text through an embedding layer to obtain a first training feature. Step S740: Perform bidirectional sequence feature extraction and context fusion processing on the first training feature through a bidirectional gated recurrent layer to obtain a second training feature. Step S750: Input the second training feature into an attention mechanism layer to obtain an attention training feature. In one embodiment, step S750 specifically includes the following steps: Perform channel attention processing on the second training feature through a channel attention module to obtain a channel attention training coefficient; Multiply the second training feature by the channel attention training coefficient to obtain a channel attention training feature; Perform spatial attention processing on the channel attention training feature through a spatial attention module to obtain a spatial attention training coefficient; Multiply the channel attention training feature by the spatial attention training coefficient to obtain an attention training feature.

[0058] Step S760: Concatenate the attention training feature and the second training feature to obtain a concatenated training feature; Step S770: Through a fully connected layer, associate the concatenated training feature with each secondary department respectively to obtain multiple associated training features; Step S780: Input the multiple associated training features into a classifier to obtain a training probability distribution of the associated training features; Step S790: Based on the training probability distribution, take the secondary department corresponding to the associated training feature with the largest probability value as the target training secondary department; It should be noted that the processing process of the training concatenated text in the registration recommendation model is the same as that of the concatenated text in the registration recommendation model.

[0059] Step S800: Calculate a loss value based on the target training secondary department and the true label of the secondary department, and iteratively update the initial registration recommendation model based on the loss value to obtain a trained registration recommendation model.

[0060] 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, and this is repeated in a loop until the number of loops reaches a preset number, so as to obtain a trained registration recommendation model. In one embodiment, the cross-entropy loss function is used as the loss function. The cross-entropy loss function is usually used for classification tasks and is applicable to scenarios where the target variable is discrete (such as the true label of the secondary department in this application).

[0061] An embodiment of the second aspect of this application provides a training method for a registration recommendation model, including the following steps: Obtain training concatenated text and the true label of the secondary department corresponding to the training concatenated text; Input the training concatenated text into an 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; Perform feature extraction processing on the training concatenated text through the embedding layer to obtain a first training feature; Performing bidirectional sequence feature extraction and context fusion processing on the first training feature through a bidirectional gated recurrent layer to obtain a second training feature; Inputting the second training feature into an attention mechanism layer to obtain an attention training feature; Concatenating the attention training feature and the second training feature to obtain a concatenated training feature; Associating the concatenated training feature with each secondary department respectively through a fully connected layer to obtain multiple associated training features; Inputting the multiple associated training features into a classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, taking the secondary department corresponding to the associated training feature with the largest probability value as the target training secondary department; Calculating a loss value based on the target training secondary department and the true label of the secondary department, and iteratively updating the initial registration recommendation model based on the loss value to obtain a trained registration recommendation model.

[0062] The third aspect embodiment of this application provides a registration recommendation device. Refer to Figure 3 , Figure 3 , which is the functional module block diagram of the registration recommendation device in the embodiment of this application. The registration recommendation device includes: An acquisition unit 310, configured to acquire voice condition information and convert the voice condition information into condition text information; A word segmentation unit 320, configured to perform word segmentation processing on the condition text information to obtain a word set, and the word set includes multiple condition words; A determination unit 330, configured to determine a target primary department from multiple primary departments based on the word set; A word concatenation unit 340, configured to concatenate each condition word in the word set and the target primary department to obtain a concatenated text; An input unit 350, configured 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; An extraction unit 360, configured to perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature; A fusion unit 370, configured to perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature; An attention unit 380, configured to input the second feature into the attention mechanism layer to obtain an attention feature; A feature concatenation unit 390, configured to concatenate the attention feature and the second feature to obtain a concatenated feature; An association unit 400 is configured to associate the spliced features with each secondary department respectively through a fully connected layer to obtain a plurality of associated features; A classification unit 410 is configured to input the plurality of associated features into a classifier to obtain a probability distribution of the associated features; A selection unit 420 is configured to, based on the probability distribution, use the secondary department corresponding to the associated feature with the largest probability value as the target secondary department; A display unit 430 is configured to display the target primary department and the target secondary department.

[0063] The registration recommendation device in the embodiments of the present application is used to execute the registration recommendation method in the embodiments of the first aspect. When executing the method, first convert the voice condition information into condition text information, perform word segmentation processing on the condition text information to obtain a set of words, determine the target primary department based on the set of words, then splice the target primary department and the set of words to obtain a spliced text, input the spliced text into the trained registration recommendation model to obtain the target secondary department. During the processing of the registration recommendation model, the target primary department can be used as prior knowledge to improve the accuracy of the registration recommendation model. The embedding layer is used to extract features from the spliced text to capture the semantic information in the spliced text; the bidirectional gated recurrent layer performs bidirectional sequence feature extraction and context fusion processing on the first feature, not only extracting the local features of the first feature, but also fusing the global context information through bidirectional processing, thereby enhancing the registration recommendation model's ability to understand and represent data. The attention mechanism layer can help the registration recommendation model understand the context information of the second feature and help the registration recommendation model select the features that contribute the most to the registration recommendation, which is beneficial to improving the accuracy of the registration recommendation model. Then, through the action of the fully connected layer and the classifier, the target secondary department is obtained, and the target primary department and the target secondary department are displayed to the user. The target secondary department is the department recommended for the user to register. In this way, the present application can automatically generate the secondary department recommended for the user to register according to the user's voice condition information, which is convenient for the user to register, and the accuracy of the recommended registration department is high.

[0064] It should be noted that the specific implementation manner of this registration recommendation device is basically the same as the specific embodiments of the above registration recommendation method, and will not be elaborated here. On the premise of meeting the requirements of the embodiments of the present application, other functional units can be set in the registration recommendation device to implement the registration recommendation method in the above embodiments.

[0065] The embodiments of the fourth aspect of the present application provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the registration recommendation method in any one of the embodiments of the first aspect. This electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0066] Refer toFigure 4 , Figure 4 is a schematic structural diagram of an electronic device according to an embodiment. The electronic device includes: A processor 401, which can be implemented in ways such as a general-purpose CPU (Central Processing Unit), a microprocessor, an 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 the present application; A memory 402, which can be implemented in forms such 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 an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 402 and are called by the processor 401 to execute the registration recommendation method of the embodiments of the present application; An input / output interface 403, which is used to implement information input and output; A communication interface 404, which is used to implement communication interaction between this device and other devices, and can achieve communication through a wired method (such as USB, network cable, etc.) or through a wireless method (such as a mobile network, WIFI, Bluetooth, etc.); A bus 405, which transmits information between various components of the device (such as the processor 401, the memory 402, the input / output interface 403, and the communication interface 404); Among them, the processor 401, the memory 402, the input / output interface 403, and the communication interface 404 achieve communication connections with each other inside the device through the bus 405.

[0067] In the embodiment of the fifth aspect of the present application, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the registration recommendation method of any one of the embodiments of the first aspect.

[0068] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include a high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0069] The embodiments described in the embodiments of the present application are for more clearly explaining the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. As can be known to those skilled in the art, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0070] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

[0071] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0072] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0073] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0074] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the mapping relationship of mapped objects and indicates that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist simultaneously. Here, A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the front and back mapped objects. "At least one (item) of the following" or its similar expressions refer to any combination of these items, including any combination of single items (items) or plural items (items). For example, at least one (item) of a, b, or c can mean: 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.

[0075] In several 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 example, the above division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of devices or units can be in electrical, mechanical, or other forms.

[0076] The units described above as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0077] In addition, in each embodiment of this application, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0078] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of this application. The foregoing storage medium includes: various media that can store programs, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0079] The preferred embodiments of the embodiments of this application have been described above with reference to the accompanying drawings, which does not limit the scope of rights of the embodiments of this application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of this application shall be within the scope of rights of the embodiments of this application.

Claims

1. A registration recommendation method, characterized in that, Including: Obtain voice condition information and convert the voice condition information into condition text information; Perform word segmentation on the condition text information to obtain a word set, where the word set includes multiple condition words; Based on the word set, determine a target first-level department from multiple first-level departments; Concatenate each of the condition words in the word set and the target first-level department to obtain a concatenated text; 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; Perform feature extraction processing on the concatenated text through the embedding layer to obtain a first feature; Perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature; Input the second feature into the attention mechanism layer to obtain an attention feature; Concatenate the attention feature and the second feature to obtain a concatenated feature; Through the fully connected layer, associate the concatenated feature with each second-level department respectively to obtain multiple associated features; Input the multiple associated features into the classifier to obtain a probability distribution of the associated features; Based on the probability distribution, use the second-level department corresponding to the associated feature with the largest probability value as the target second-level department; Display the target first-level department and the target second-level department.

2. The registration recommendation method according to claim 1, wherein 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 an attention feature includes: Perform channel attention processing on the second feature through the channel attention module to obtain a channel attention coefficient; Multiply the second feature by the channel attention coefficient to obtain a channel attention feature; Perform spatial attention processing on the channel attention feature through the spatial attention module to obtain a spatial attention coefficient; Multiply the channel attention feature by the spatial attention coefficient to obtain the attention feature.

3. The registration recommendation method according to claim 1, wherein The step of determining a target first-level department from multiple first-level departments based on the word set includes: Match the condition words in the word set with multiple preset first-level department keyword corpora to obtain multiple matching degree values; wherein, the first-level department keyword corpora correspond to the first-level departments one by one; Use the first-level department keyword corpus corresponding to the largest value among the multiple matching degree values as the target keyword corpus, and based on the target keyword corpus, determine the target first-level department from multiple first-level departments.

4. The registration recommendation method according to claim 1, wherein Before the step of displaying the target first-level department and the target second-level department, it further includes: Detect the target second-level department and the target first-level department; In the case where it is detected that the target second-level department does not belong to the lower classification of the target first-level department, re-perform word segmentation on the condition text information to obtain a new word set, and jump to the step of determining a target first-level department from multiple first-level departments based on the word set.

5. The registration recommendation method according to claim 1, wherein After displaying the target first-level department and the target second-level department, it further includes: Displaying a registration department approval button; In response to a first operation instruction for the registration department approval button, displaying an identity information text box; In response to a second operation instruction for the identity information text box, obtaining the input text in the identity information text box; Sending the input text, the target first-level department, and the target second-level department to a registration system for registration.

6. The registration recommendation method according to claim 1, wherein The registration recommendation model is obtained through the following steps: Obtaining training concatenated text and the true label of the second-level department corresponding to the training concatenated text; Inputting the training concatenated text into an initial registration recommendation model; Performing feature extraction processing on the training concatenated text through the embedding layer to obtain a first training feature; Performing bidirectional sequence feature extraction and context fusion processing on the first training feature through the bidirectional gated recurrent layer to obtain a second training feature; Inputting the second training feature into the attention mechanism layer to obtain an attention training feature; Concatenating the attention training feature and the second training feature to obtain a concatenated training feature; Through the fully connected layer, associating the concatenated training feature with each of the second-level departments respectively to obtain multiple associated training features; Inputting the multiple associated training features into the classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, taking the second-level department corresponding to the associated training feature with the largest probability value as the target training second-level department; Calculating a loss value based on the target training second-level department and the true label of the second-level department, and iteratively updating the initial registration recommendation model based on the loss value to obtain the trained registration recommendation model.

7. A training method for a registration recommendation model, characterized in that, It includes: Obtaining training concatenated text and the true label of the second-level department corresponding to the training concatenated text; Inputting the training concatenated text into an 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; Performing feature extraction processing on the training concatenated text through the embedding layer to obtain a first training feature; Performing bidirectional sequence feature extraction and context fusion processing on the first training feature through the bidirectional gated recurrent layer to obtain a second training feature; Inputting the second training feature into the attention mechanism layer to obtain an attention training feature; Concatenating the attention training feature and the second training feature to obtain a concatenated training feature; Through the fully connected layer, associating the concatenated training feature with each of the second-level departments respectively to obtain multiple associated training features; Inputting the multiple associated training features into the classifier to obtain the training probability distribution of the associated training features; Based on the training probability distribution, taking the second-level department corresponding to the associated training feature with the largest probability value as the target training second-level department; Calculate the loss value based on the target training secondary department and the true label of the secondary department, and iteratively update the initial registration recommendation model based on the loss value to obtain the trained registration recommendation model.

8. A registration recommendation device, characterized in that, It includes: An acquisition unit, configured to acquire voice condition information and convert the voice condition information into condition text information; A word segmentation unit, configured to perform word segmentation processing on the condition text information to obtain a word set, where the word set includes multiple condition words; A determination unit, configured to determine a target primary department from multiple primary departments based on the word set; A word splicing unit, configured to splice each condition word in the word set and the target primary department to obtain a spliced text; An input unit, configured to input the spliced 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; An extraction unit, configured to perform feature extraction processing on the spliced text through the embedding layer to obtain a first feature; A fusion unit, configured to perform bidirectional sequence feature extraction and context fusion processing on the first feature through the bidirectional gated recurrent layer to obtain a second feature; An attention unit, configured to input the second feature into the attention mechanism layer to obtain an attention feature; A feature splicing unit, configured to splice the attention feature and the second feature to obtain a spliced feature; An association unit, configured to associate the spliced feature with each secondary department respectively through the fully connected layer to obtain multiple associated features; A classification unit, configured to input the multiple associated features into the classifier to obtain a probability distribution of the associated features; A selection unit, configured to, based on the probability distribution, use the secondary department corresponding to the associated feature with the largest probability value as the target secondary department; A display unit, configured to display the target primary department and the target secondary department.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the registration recommendation method according to any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the registration recommendation method according to any one of claims 1 to 6.

Images