Medical triage method, device and equipment and storage medium

Through emotion recognition and department classification network model, the respondents are automatically matched, and the problems of low efficiency and accuracy of medical triage in the existing technology are solved, and the efficiency and user experience of online consultation are improved.

CN118942668APending Publication Date: 2024-11-12BEIJING JINGDONG TUOXIAN TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310524725.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-10
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the prior art, patients cannot accurately match the respondent before conducting online consultation, resulting in low medical triage efficiency and accuracy, and manual triage is time-consuming and labor-intensive, reducing user experience.

Method used

By obtaining the patient's complaint information, using the pre-trained emotion recognition network model and department classification network model, the patient's emotional information and matching department are automatically determined, and matching the patient's emotional information based on the personality information of the subject to the diagnosis, selecting the most suitable one the prescription for diagnosis.

Benefits of technology

It realizes automated medical triage, improves the efficiency and accuracy of medical triage, takes into account the emotional factors of patients, provides a more matching treatment and improves the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942668A_ABST
    Figure CN118942668A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a medical triage method and device, equipment and a storage medium. The method comprises the following steps: acquiring target chief complaint information of a target doctor seeing party; inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining target emotion information of the target doctor seeing party based on output of the emotion recognition network model; inputting the target chief complaint information into a department classification network model obtained by pre-training, and determining a target department matched with the target doctor seeing party based on the output of the department classification network model; matching the character information of each to-be-selected consultant in the target department with the target emotion information, and determining a first matching degree between each to-be-selected consultant and the target doctor seeing party; and based on the first matching degree corresponding to each to-be-selected patient applicant, determining a target patient applicant matched with the target patient seeing party. Through the technical scheme of the embodiment of the invention, automatic triage can be realized, and the efficiency and accuracy of medical triage are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of Internet medical technology, and in particular to a medical triage method, device, equipment and storage medium. Background Art

[0002] With the rapid development of computer technology and Internet medicine, it has become popular for people to consult doctors through online consultation. Before online consultation, medical triage is required so that the matching consultation party can be used to conduct online consultation for the consultation party.

[0003] At present, the patient usually chooses the patient to be consulted by the patient himself. However, in the process of realizing the present disclosure, the inventor found that there are at least the following problems in the prior art:

[0004] Since the patient himself may not be aware of the lesions that cause the symptoms, multiple referrals are required to find a matching patient, which reduces the efficiency and accuracy of medical triage. In addition, manual triage is time-consuming and labor-intensive, reducing the user experience. Summary of the invention

[0005] The embodiments of the present disclosure provide a medical triage method, apparatus, device and storage medium to achieve automatic triage and improve the efficiency and accuracy of medical triage.

[0006] In a first aspect, an embodiment of the present disclosure provides a medical triage method, comprising:

[0007] Obtain target chief complaint information of target patients;

[0008] Inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model;

[0009] Inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on an output of the department classification network model;

[0010] Matching the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient;

[0011] Based on the first matching degree corresponding to each candidate patient, a target patient matching the target patient is determined.

[0012] In a second aspect, the present disclosure also provides a medical triage device, including:

[0013] A target chief complaint information acquisition module is used to acquire the target chief complaint information of the target patient;

[0014] A target emotion information determination module, used to input the target chief complaint information into a pre-trained emotion recognition network model, and determine the target emotion information of the target patient based on the output of the emotion recognition network model;

[0015] A target department determination module, used for inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on the output of the department classification network model;

[0016] A first matching degree determination module is used to match the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient;

[0017] The target patient determination module is used to determine a target patient matching the target patient based on the first matching degree corresponding to each candidate patient.

[0018] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

[0019] one or more processors;

[0020] A memory for storing one or more programs;

[0021] When the one or more programs are executed by the one or more processors, the one or more processors implement the medical triage method provided by any embodiment of the present disclosure.

[0022] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a medical triage method as provided in any embodiment of the present disclosure.

[0023] One embodiment of the above invention has the following advantages or beneficial effects:

[0024] By inputting the target complaint information of the target patient into the pre-trained emotion recognition network model, the target emotion information of the target patient can be automatically and accurately determined. By inputting the target complaint information into the pre-trained department classification network model, the target department matching the target patient can be automatically and accurately determined. The personality information of each candidate patient in the target department is matched with the target emotion information, and the first matching degree between each candidate patient and the target patient is determined. Based on the first matching degree corresponding to each candidate patient, the target patient matching the target patient is determined, thereby realizing automatic triage. In addition, the influence of the patient's emotions on the triage results is taken into account, and a more matching target patient can be obtained. For example, if the patient is currently in a more anxious mood, a gentle patient is assigned for consultation, which improves the efficiency and accuracy of medical triage and also improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 is a flow chart of a medical triage method provided by an embodiment of the present disclosure;

[0027] Figure 2 is a flow chart of another medical triage method provided by an embodiment of the present disclosure;

[0028] Figure 3 is an example diagram of a medical triage process involved in an embodiment of the present disclosure;

[0029] Figure 4 is a flow chart of another medical triage method provided by an embodiment of the present disclosure;

[0030] Figure 5 is a structural schematic diagram of a medical triage device provided by an embodiment of the present disclosure;

[0031] Figure 6 It is a structural schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0032] The present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present disclosure, rather than to limit the present disclosure. It should also be noted that, for ease of description, only parts related to the present disclosure are shown in the accompanying drawings, rather than all structures.

[0033] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0034] Figure 1 This is a flowchart of a medical triage method provided by an embodiment of the present disclosure. This embodiment is applicable to online medical triage. The method can be performed by a medical triage device, which can be implemented by software and / or hardware and integrated into an electronic device. Figure 1 As shown, the method specifically comprises the following steps:

[0035] S110, obtaining target chief complaint information of the target patient.

[0036] The target patient may refer to a patient who currently needs to see a doctor. For example, the target patient may refer to a patient who seeks medical treatment online. The target complaint information may be a description of the symptoms reported by the target patient. The target complaint information may be represented in text form.

[0037] Specifically, the target chief complaint information input by the target patient during online consultation can be obtained.

[0038] S120, inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model.

[0039] The emotion recognition network model may be a neural network model for identifying the patient's emotional information. The emotion recognition network model may be a classification network model, that is, emotion recognition is performed by predicting the probability value of the patient's emotional information being each type of classified emotional information. The emotion recognition network model may be pre-trained based on the sample chief complaint information of the sample patient and the sample patient's emotional information label. The target emotional information of the target patient may include, but is not limited to, at least one of: calm, anxious, afraid, irritable, and fearful.

[0040] Specifically, the emotion recognition network model obtained through pre-training can accurately recognize emotions based on the input target complaint information, and output the predicted probability value of the patient's emotion information being each type of emotion information. Based on the predicted probability value output by the emotion recognition network model, the emotion information with the largest predicted probability value can be determined as the target emotion information of the target patient. Emotion information with a predicted probability value greater than a preset threshold can also be determined as the target emotion information of the target patient. If the determined target emotion information includes at least two types of emotion information, it can be matched separately based on each type of emotion information, so as to consider the impact of multiple types of emotion information on the triage results, and further improve the accuracy of triage.

[0041] S130, inputting the target chief complaint information into the pre-trained department classification network model, and determining the target department matching the target patient based on the output of the department classification network model.

[0042] The department classification network model may be a classification network model for triage departments, that is, department matching is performed by predicting the probability value of each department that the department matching the patient is a patient. The department classification network model may be obtained by pre-training based on sample chief complaint information of sample patients and department labels matching the sample patients.

[0043] Specifically, the department classification network model obtained by pre-training can accurately match departments based on the input target chief complaint information, and output the predicted probability value of each department that matches the patient. Based on the predicted probability value output by the department classification network model, the department with the largest predicted probability value can be determined as the target department that matches the target patient. The department classification network model can also directly output the target department that matches the target patient. By using the department classification network model obtained by pre-training, departments can be automatically triaged and the accuracy of the triaged departments can be guaranteed.

[0044] It should be noted that this embodiment does not limit the execution order of step S120 and step S130. For example, step S120 may be executed first and then step S130, or step S130 may be executed first and then step S120, or step S120 and step S130 may be executed simultaneously.

[0045] S140, matching the personality information of each candidate patient in the target department with the target emotion information, and determining a first matching degree between each candidate patient and the target patient.

[0046] Among them, the target department may include multiple candidates for consultation. The personality information of the candidate for consultation can be used to characterize the consultation style of the candidate for consultation. For example, the personality information of the candidate for consultation may include but is not limited to: at least one of gentleness, patience and rigor. The personality information of the candidate for consultation can be determined in advance based on the historical dialogue information corresponding to the candidate for consultation. The first matching degree can be used to characterize the degree to which the personality of the candidate for consultation matches the emotion of the candidate for consultation.

[0047] Specifically, a personality and emotion matching dictionary can be established, which includes the matching degree between each personality information of the patient and each emotion information of the patient. The matching degree can be divided into: the matching degree corresponding to a very good match, such as 1, the matching degree corresponding to a relatively good match, such as 0.8, the matching degree corresponding to a general match, such as 0.6, the matching degree corresponding to a not very good match is 0.4, and the matching degree corresponding to a completely unmatched match is 0.2. For example, the matching degree between a gentle personality and an anxious emotion is 1, and the matching degree between a rigorous personality and an anxious emotion is 0.6. Based on the personality and emotion matching dictionary, the first matching degree between the personality information of each candidate patient in the target department and the target emotion information can be determined. If there is only one target emotion information for the target patient, the matching degree between the target emotion information and the personality information of each candidate patient is directly determined as the first matching degree. If the target patient has at least two target emotion information, the matching degree between each target emotion information and the personality information of each candidate patient is determined, and the matching degree corresponding to each candidate patient under various target emotion information is averaged, and the obtained average matching degree is determined as the first matching degree between each candidate patient and the target patient.

[0048] S150: Determine a target patient that matches the target patient based on the first matching degree corresponding to each candidate patient.

[0049] Specifically, the candidate for consultation with the highest first matching degree in the target department can be determined as the target consultation party that matches the target consultation party. The candidate for consultation with a first matching degree greater than a preset matching degree can also be used as the target consultation party. The first preset number of candidate for consultation with the highest first matching degree can also be used as the target consultation party, such as the first 200 candidate for consultation with the highest first matching degree as the target consultation party. If there are multiple target consultation parties, the introduction information of each target consultation party can be displayed to the target consultation party, so that the target consultation party can independently select a target consultation party based on the introduction information, and the selected target consultation party provides consultation services for the target consultation party. Alternatively, a consultation request can also be sent to all target consultation parties, and the target consultation party with the fastest consultation can provide consultation services for the target consultation party by grabbing the order. By triaging the consultation party based on the mood of the consultation party after the department triage, the consultation party that matches the mood of the consultation party can be used for consultation, thereby improving the consultation experience of the consultation party under the premise of ensuring the accuracy of triage.

[0050] The technical solution of this embodiment can automatically and accurately determine the target emotional information of the target patient by inputting the target complaint information of the target patient into the pre-trained emotion recognition network model, and can automatically and accurately determine the target department that matches the target patient by inputting the target complaint information into the pre-trained department classification network model. The personality information of each candidate patient in the target department is matched with the target emotional information, and the first matching degree between each candidate patient and the target patient is determined, and based on the first matching degree corresponding to each candidate patient, the target patient that matches the target patient is determined, thereby realizing automatic triage, and taking into account the influence of the patient's emotions on the triage results, a more matching target patient can be obtained, which improves the efficiency and accuracy of medical triage, and also improves the user experience.

[0051] On the basis of the above technical solution, S150 may include: obtaining the first parameter information corresponding to the target patient under the preset matching parameters and the second parameter information corresponding to each selected patient; determining the second matching degree between each selected patient and the target patient under the preset matching parameters based on the first parameter information and the second parameter information; determining the target patient that matches the target patient based on the first matching degree and the second matching degree corresponding to each selected patient.

[0052] The preset matching parameters may be matching parameters that are pre-set based on business needs and that can affect the accuracy of triage. For example, the preset matching parameters may include, but are not limited to, at least one of gender, age, and region. The second matching degree may refer to the matching degree between the candidate patient and the target patient under the same preset matching parameters. For example, the second matching degree may refer to gender matching degree, age matching degree, or region matching degree.

[0053] Specifically, based on the first parameter information corresponding to the target patient under the same preset matching parameter and the second parameter information corresponding to each candidate patient, the second matching degree between each candidate patient and the target patient under the preset matching parameter can be determined. For example, the first characterization vector corresponding to the first parameter information and the second characterization vector corresponding to the second parameter information can be determined, and the Euclidean distance between the first parameter information and the second parameter information can be determined based on the first characterization vector and the second characterization vector, and the Euclidean distance can be used as the second matching degree between the candidate patient and the target patient. The first matching degree and the second matching degree corresponding to each preset matching parameter are weighted and summed to obtain the target matching degree corresponding to each candidate patient, and the candidate patient with the largest target matching degree can be determined as the target patient matching the target patient; the candidate patient with a target matching degree greater than the preset matching degree can also be used as the target patient; the first preset number of candidate patients with the highest target matching degree can also be used as the target patient. By triaging patients based on both patient emotions and preset matching parameters, more matched patients can be obtained, thereby further improving the accuracy of medical triage and enhancing the user experience.

[0054] Figure 2 A flowchart of another medical triage method provided for an embodiment of the present disclosure. In this embodiment, based on the above embodiments, the emotion recognition network model includes: a first word vector generation sub-model, an emotion feature extraction sub-model and a first classification sub-model, and the department classification network model includes: at least one feature extraction sub-model and a second classification sub-model. On this basis, the medical triage process is described in detail. The explanations of the terms that are the same or corresponding to the above embodiments are not repeated here.

[0055] See also Figure 2 Another medical triage method provided in this embodiment specifically includes the following steps:

[0056] S210, obtaining target chief complaint information of the target patient.

[0057] Specifically, before inputting the model, the target complaint information can be preprocessed to reduce data noise. For example, stop words that are not related to emotions in the target complaint information are filtered. Stop words include English characters, numbers, mathematical symbols, punctuation marks, and frequently used Chinese characters or words. On the one hand, these stop words are frequently used in the text and take up a lot of storage space. On the other hand, these stop words have a low degree of relevance to the content to be expressed by the complaint information and have no practical meaning. Therefore, these stop words need to be removed during preprocessing.

[0058] S220, input the target complaint information into the first word vector generation sub-model in the emotion recognition network model for text conversion to generate a first word vector sequence.

[0059] The first word vector generation submodel may be a network model Word2Vec for generating word vectors. For example, the first word vector generation submodel may be, but is not limited to, a skip-gram model or a continuous bag of words model CBOW (Continuous Bag Of Words). Skip-gram generates word vectors by predicting context words from words in a window. CBOW generates word vectors by predicting words between contexts based on context words. The first word vector sequence may include a word vector corresponding to each word in the target complaint information.

[0060] Specifically, Figure 3 An example diagram of a medical triage process is given. Figure 3 As shown, the preprocessed target complaint information can be input into the first word vector generation sub-model in the emotion recognition network model. The first word vector generation sub-model can encode each word in the input target complaint information, generate a corresponding first word vector, and output the obtained first word vector sequence. Therefore, based on the output of the first word vector generation sub-model, the first word vector sequence corresponding to the text of the target complaint information can be obtained.

[0061] S230: Input the first word vector sequence into the emotion feature extraction sub-model in the emotion recognition network model to extract context emotion features, and obtain extracted target emotion feature information.

[0062] The emotion feature extraction sub-model may be a neural network model for extracting context emotion features in target complaint information. For example, the emotion feature extraction sub-model may be, but is not limited to, a gated recurrent unit (GRU).

[0063] Specifically, Figure 3 As shown, the first word vector sequence can be input into the emotion feature extraction sub-model in the emotion recognition network model. The emotion feature extraction sub-model extracts contextual emotional features from the input first word vector sequence, and outputs the extracted target emotion feature information. Thus, based on the output of the emotion feature extraction sub-model, the target emotion feature information in the text of the target complaint information can be obtained.

[0064] Exemplarily, the emotion feature extraction sub-model may include: a forward gated recurrent unit, a reverse gated recurrent unit and a feature splicing unit. Accordingly, S230 may include: inputting the first word vector sequence into the forward gated recurrent unit to perform emotion feature extraction of the forward sequence to obtain the extracted first emotion feature information; inputting the first word vector sequence into the reverse gated recurrent unit to perform emotion feature extraction of the reverse sequence to obtain the extracted second emotion feature information; inputting the first emotion feature information and the second emotion feature information into the feature splicing unit for splicing to obtain the target emotion feature information.

[0065] Specifically, the emotion feature extraction sub-model can use two memory units, namely, a forward gated recurrent unit and a reverse gated recurrent unit, to process the forward sequence and the reverse sequence respectively, and concatenate the first emotion feature information and the second emotion feature information output by the forward gated recurrent unit and the reverse gated recurrent unit respectively as the target emotion feature information finally extracted. By using the forward gated recurrent unit and the reverse gated recurrent unit, bidirectional feature extraction can be performed, thereby fully extracting contextual emotion features and further improving the accuracy of emotion recognition.

[0066] S240, inputting the target emotion feature information into the first classification sub-model in the emotion recognition network model to perform emotion classification, obtaining the predicted probability value corresponding to each emotion information and outputting it.

[0067] The first classification sub-model may be a classifier for classifying emotional information. For example, the first classification sub-model may be, but is not limited to, a multilayer perceptron (MLP). Specifically, Figure 3 As shown, the target emotion feature information can be input into the first classification sub-model in the emotion recognition network model. The first classification sub-model performs classification processing based on the input target emotion feature information, obtains the predicted probability value corresponding to each emotion information, and outputs it.

[0068] For example, when training the emotion recognition network model, in order to strike a balance between the update speed and the number of updates of the network weights, the Mini-batch gradient descent strategy can be used, such as setting the batch-size to 128, to achieve the effect of increasing the learning speed and ensuring the convergence of the model. In order to prevent the occurrence of overfitting during the training of the emotion recognition network model, a random deactivation Dropout strategy can be added to the emotion recognition network model to improve the generalization ability of the model by abandoning the modification of some weights, where the discard rate in the Dropout strategy can be set to 0.5.

[0069] S250: Determine target emotion information of the target patient based on the output of the first classification sub-model in the emotion recognition network model.

[0070] Specifically, Figure 3 As shown, based on the predicted probability value corresponding to each type of emotion information output by the first classification sub-model in the emotion recognition network model, the emotion information with the largest predicted probability value can be determined as the target emotion information of the target patient. Alternatively, the emotion information with a predicted probability value greater than a preset threshold can be determined as the target emotion information of the target patient.

[0071] S260, input the target chief complaint information into each feature extraction sub-model in the department classification network model to extract corresponding dimensional features, and obtain the extracted chief complaint feature information.

[0072] The feature extraction submodel may be a model for extracting feature information from a certain feature dimension in the target main complaint information. The number of feature extraction submodels may be one or more. For example, the feature extraction submodel may include at least one of a word frequency-inverse text frequency statistics submodel, a second word vector generation submodel, a document topic generation submodel, and a part of speech extraction submodel.

[0073] Among them, the term frequency-inverse text frequency statistical sub-model can be used to extract the importance of each word. If a word appears frequently in a document and rarely appears in other documents in the corpus, it is considered that this word has a high representation ability for this document. Term frequency TF (Term Frequency) is the frequency of a word appearing in the target complaint information, which can be normalized by dividing the term frequency by the total number of words. Inverse document frequency IDF (Inverse Document Frequency) refers to the frequency of a word appearing in other documents. The total number of documents is divided by the number of documents containing the word, and then the logarithm of the division result is taken to obtain the inverse text frequency. By inputting the target complaint information into the term frequency-inverse text frequency statistical sub-model for feature extraction under the importance dimension, the extracted complaint feature information used to characterize the importance of each word is obtained.

[0074] Among them, the second word vector generation submodel can be a network model Word2Vec for generating word vectors. For example, the first word vector generation submodel can be but not limited to: a skip-gram model or a continuous bag of words model CBOW (Continuous Bag Of Words). The target main complaint information is input into the second word vector generation submodel to perform feature extraction under the word vector dimension to obtain an extracted second word vector sequence. The second word vector sequence can be used as the extracted main complaint feature information, or after determining the second word vector sequence, sliding windows of sizes 2, 3 and 4 can be used to slide on the second word vector sequence to determine the maximum value and average value of the word vector in each sliding window. For the classification model, the greatest contribution is not necessarily the entire sentence, but may be a part of words and phrases, so using a sliding window to extract word vectors of several words helps to extract more effective features. The extracted second word vector sequence and the maximum value and average value of the word vector determined by sliding can be used as the extracted main complaint feature information.

[0075] Among them, the document topic generation submodel can be a network model for predicting the topic distribution of a document. For example, the document topic generation submodel can be but is not limited to: LDA (Latent Diriclet Allocation) topic model. By inputting the target main complaint information into the document topic generation submodel to perform feature extraction under the topic dimension, the main complaint feature information extracted for characterizing the topic similarity is obtained. For example, including 30 topics, the extracted main complaint feature information is a 30-dimensional feature vector, and each element value in the vector represents the matching degree with the corresponding topic.

[0076] The part-of-speech extraction submodel may be a network model for extracting part-of-speech information. Parts of speech may include nouns, adjectives, and verbs. For example, the part-of-speech extraction submodel may be, but is not limited to, a named entity recognition model. By inputting the target main complaint information into the part-of-speech extraction submodel to perform feature extraction under the part-of-speech dimension, main complaint feature information such as part-of-speech information, number or proportion of nouns, number or proportion of adjectives, and number or proportion of adjectives are extracted.

[0077] S270, input the chief complaint feature information into the second classification sub-model in the department classification network model to perform department classification, obtain the predicted probability value corresponding to each department and output it.

[0078] The second classification sub-model may be a classifier for classifying departments. For example, the second classification sub-model may also be, but not limited to, a multilayer perceptron (MLP). Specifically, Figure 3As shown, the chief complaint feature information output by the feature extraction sub-model can be input into the second classification sub-model, and the second classification sub-model performs department classification based on the input chief complaint feature information, obtains the prediction probability value corresponding to each department and outputs it. If there are at least two feature extraction sub-models, the chief complaint feature information extracted by each feature extraction sub-model can be spliced ​​and merged, and the merged chief complaint feature information can be input into the second classification sub-model for department classification, thereby using feature information of many dimensions to classify departments, thereby improving the accuracy of department classification.

[0079] S280. Determine a target department that matches the target patient based on the output of the second classification sub-model in the department classification network model.

[0080] Specifically, based on the predicted probability value corresponding to each department output by the second classification sub-model in the department classification network model, the department with the largest predicted probability value can be determined as the target department that matches the target patient.

[0081] S290. Match the personality information of each candidate patient in the target department with the target emotional information to determine a first matching degree between each candidate patient and the target patient.

[0082] Specifically, Figure 3 As shown, the personality and emotion of the patient are matched based on the target department determined by the department classification network model and the target emotion information determined by the emotion recognition network model to obtain the first matching degree between each candidate patient and the target patient.

[0083] S291. Determine a target patient that matches the target patient based on the first matching degree corresponding to each candidate patient.

[0084] The technical solution of this embodiment can more accurately perform emotion recognition by utilizing an emotion recognition network model including a first word vector generation sub-model, an emotion feature extraction sub-model and a first classification sub-model, and can more accurately perform department classification by utilizing a department classification network model including at least one feature extraction sub-model and a second classification sub-model, thereby further improving the accuracy of medical triage.

[0085] Figure 4 This is a flowchart of another medical triage method provided by an embodiment of the present disclosure. Based on the above embodiments, this embodiment describes in detail the process of determining the personality information of the diagnosis party. The explanations of the terms that are the same or corresponding to the above embodiments are not repeated here.

[0086] See also Figure 4 Another medical triage method provided in this embodiment specifically includes the following steps:

[0087] S410. Determine the valid reply information corresponding to each patient in each department based on the historical conversation information corresponding to each patient in each department.

[0088] The historical dialogue information may be the information of the consultation dialogue between the patient and the patient in the recent period. The reply information may refer to the information of the patient's reply to the patient's question. The effective reply information refers to the reply information that can reflect the patient's personality.

[0089] Specifically, for each patient in each department, historical conversation information corresponding to the patient can be obtained based on the historical conversation log, and effective information can be extracted from the historical conversation information to obtain effective reply information corresponding to the patient.

[0090] Exemplarily, S410 may include: extracting historical reply information from the historical conversation information corresponding to each patient in each department; deleting the drug names in the historical reply information, and using the deleted historical reply information whose length is greater than a preset length as valid reply information.

[0091] Specifically, for each patient in each department, all historical reply information corresponding to the patient's historical conversation information can be extracted, and the drug name in each historical reply information can be deleted. The length of each historical reply information after deleting the drug name will be counted, and the historical reply information with a length greater than a preset length after deleting the drug name will be taken as valid reply information. For example, the historical reply information with a reply number of more than 10 words after deleting the drug name will be taken as valid reply information, so as to improve the accuracy of the patient's personality recognition.

[0092] S420, input the valid response information into the pre-trained personality recognition network model, and determine the personality information corresponding to each patient in each department based on the output of the personality recognition network model.

[0093] The personality recognition network model may be a neural network model for identifying the personality information of the patient. The personality recognition network model may be a classification network model, that is, the personality recognition is performed by predicting the probability value of the patient's personality information as each type of personality information after classification. The personality recognition network model may be pre-trained based on the effective response information of the sample patient and the emotional information label of the sample patient. The personality information corresponding to the patient may include but is not limited to at least one of: gentle, patient and rigorous.

[0094] Specifically, for each patient in each department, the pre-trained personality recognition network model can accurately recognize the personality based on the input valid response information, and output the predicted probability value of the patient's personality information being each type of personality information. Based on the predicted probability value output by the personality recognition network model, the personality information with the largest predicted probability value can be used to determine the personality information corresponding to the patient.

[0095] Exemplarily, the personality recognition network model includes: a document vector generation sub-model and a third classification sub-model. The document vector generation sub-model may be a network model for generating document vectors, such as a Doc2Vec model. The third classification sub-model may be a classifier for classifying personality. For example, the third classification sub-model may be, but is not limited to, a convolutional neural network for implementing a classification function.

[0096] Exemplarily, "inputting the valid response information into the pre-trained personality recognition network model" in S420 may include: inputting the valid response information into the document vector generation sub-model in the personality recognition network model to generate a document vector, and inputting the generated document vector into the third classification sub-model in the personality recognition network model to perform personality recognition, and outputting the personality information of the patient as the predicted probability value of each personality information. By using the document vector generation sub-model and the third classification sub-model to more accurately perform personality recognition, the accuracy of medical triage is further improved.

[0097] S430, obtaining target chief complaint information of the target patient.

[0098] S440, inputting the target chief complaint information into the pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model.

[0099] S450, input the target chief complaint information into the pre-trained department classification network model, and determine the target department that matches the target patient based on the output of the department classification network model.

[0100] S460: Match the personality information of each candidate patient in the target department with the target emotional information to determine a first matching degree between each candidate patient and the target patient.

[0101] Specifically, based on the pre-identified personality information of each patient, the personality information of each candidate patient in the target department can be quickly obtained, thereby further improving the efficiency of medical triage.

[0102] S470: Determine a target patient that matches the target patient based on the first matching degree corresponding to each candidate patient.

[0103] The technical solution of this embodiment can accurately determine the personality information corresponding to each patient in each department by utilizing the personality recognition network model obtained through pre-training, thereby ensuring the accuracy of medical triage.

[0104] The following is an embodiment of a medical triage device provided in an embodiment of the present disclosure. The device and the medical triage methods of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the medical triage device, reference can be made to the embodiments of the above-mentioned medical triage methods.

[0105] Figure 5 This is a schematic diagram of the structure of a medical triage device provided in an embodiment of the present disclosure. This embodiment can be applied to online medical triage. Figure 5 As shown, the device specifically includes: a target chief complaint information acquisition module 510, a target emotion information determination module 520, a target department determination module 530, a first matching degree determination module 540 and a target patient determination module 550.

[0106] Among them, the target chief complaint information acquisition module 510 is used to obtain the target chief complaint information of the target patient; the target emotion information determination module 520 is used to input the target chief complaint information into the pre-trained emotion recognition network model, and determine the target emotion information of the target patient based on the output of the emotion recognition network model; the target department determination module 530 is used to input the target chief complaint information into the pre-trained department classification network model, and determine the target department matching the target patient based on the output of the department classification network model; the first matching degree determination module 540 is used to match the personality information of each candidate patient in the target department with the target emotion information, and determine the first matching degree between each candidate patient and the target patient; the target patient determination module 550 is used to determine the target patient matching the target patient based on the first matching degree corresponding to each candidate patient.

[0107] The technical solution of this embodiment can automatically and accurately determine the target emotional information of the target patient by inputting the target complaint information of the target patient into the pre-trained emotion recognition network model, and can automatically and accurately determine the target department that matches the target patient by inputting the target complaint information into the pre-trained department classification network model. The personality information of each candidate patient in the target department is matched with the target emotional information, and the first matching degree between each candidate patient and the target patient is determined, and based on the first matching degree corresponding to each candidate patient, the target patient that matches the target patient is determined, thereby realizing automatic triage, and taking into account the influence of the patient's emotions on the triage results, a more matching target patient can be obtained. For example, if the patient is currently in a more anxious mood, a gentle patient is assigned for consultation, which improves the efficiency and accuracy of medical triage, and also improves the user experience.

[0108] Optionally, the emotion recognition network model includes: a first word vector generation sub-model, an emotion feature extraction sub-model and a first classification sub-model;

[0109] The target emotion information determination module 520 includes:

[0110] A first word vector sequence generation submodule, used for inputting the target chief complaint information into a first word vector generation submodel for text conversion to generate a first word vector sequence;

[0111] An emotion feature information extraction submodule, used for inputting the first word vector sequence into the emotion feature extraction submodel to extract context emotion features and obtain extracted target emotion feature information;

[0112] The predicted probability value output submodule is used to input the target emotion feature information into the first classification submodel for emotion classification, obtain the predicted probability value corresponding to each emotion information and output it.

[0113] Optionally, the emotion feature extraction sub-model includes: a forward gated recurrent unit, a reverse gated recurrent unit and a feature concatenation unit;

[0114] The emotion feature information extraction submodule is specifically used for:

[0115] The first word vector sequence is input into the forward gated recurrent unit to extract the sentiment features of the forward sequence, and the extracted first sentiment feature information is obtained; the first word vector sequence is input into the reverse gated recurrent unit to extract the sentiment features of the reverse sequence, and the extracted second sentiment feature information is obtained; the first sentiment feature information and the second sentiment feature information are input into the feature splicing unit for splicing, and the target sentiment feature information is obtained.

[0116] Optionally, the department classification network model includes: at least one feature extraction sub-model and a second classification sub-model;

[0117] The target department determination module 530 is specifically used to: input the target chief complaint information into each of the feature extraction sub-models to extract the corresponding dimensional features, and obtain the extracted chief complaint feature information; input the chief complaint feature information into the second classification sub-model to classify the department, obtain the corresponding predicted probability value of each department and output it.

[0118] Optionally, the feature extraction submodel includes: at least one of a word frequency-inverse text frequency statistics submodel, a second word vector generation submodel, a document topic generation submodel and a part-of-speech extraction submodel.

[0119] Optionally, the device further comprises:

[0120] An effective response information determination module, used to determine the effective response information corresponding to each patient in each department based on the historical conversation information corresponding to each patient in each department;

[0121] The personality recognition module is used to input the effective response information into a pre-trained personality recognition network model, and determine the personality information corresponding to each patient in each department based on the output of the personality recognition network model.

[0122] Optionally, the effective reply information determination module is specifically used to:

[0123] Extract the historical reply information from the historical conversation information corresponding to each patient in each department; delete the drug names in the historical reply information, and use the deleted historical reply information whose length is greater than a preset length as valid reply information.

[0124] Optionally, the target patient determination module 550 is specifically configured to:

[0125] Obtain the first parameter information corresponding to the target patient under the preset matching parameters and the second parameter information corresponding to each candidate patient; determine the second matching degree between each candidate patient and the target patient under the preset matching parameters based on the first parameter information and the second parameter information; determine the target patient that matches the target patient based on the first matching degree and the second matching degree corresponding to each candidate patient.

[0126] The medical triage device provided in the embodiments of the present disclosure can execute the medical triage method provided in any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects for executing the medical triage method.

[0127] It is worth noting that in the embodiment of the above-mentioned medical triage device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present disclosure.

[0128] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. Figure 6 A block diagram of an exemplary electronic device 12 suitable for use in implementing embodiments of the present disclosure is shown. Figure 6 The electronic device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0129] like Figure 6 As shown, the electronic device 12 is in the form of a general purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0130] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0131] The electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0132] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 6 not shown, usually called a "hard drive"). Although Figure 6Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present disclosure.

[0133] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods of the embodiments described in the present disclosure.

[0134] The electronic device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), may communicate with one or more devices that enable a user to interact with the electronic device 12, and / or may communicate with any device that enables the electronic device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. Furthermore, the electronic device 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0135] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, for example, implementing a medical triage method provided by the embodiment of the present invention, the method comprising:

[0136] Obtain target chief complaint information of target patients;

[0137] Inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model;

[0138] Inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on an output of the department classification network model;

[0139] Matching the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient;

[0140] Based on the first matching degree corresponding to each candidate patient, a target patient matching the target patient is determined.

[0141] Of course, those skilled in the art can understand that the processor can also implement the technical solution of the medical triage method provided in any embodiment of the present disclosure.

[0142] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the medical triage method provided in any embodiment of the present disclosure are implemented. The method includes:

[0143] Obtain target chief complaint information of target patients;

[0144] Inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model;

[0145] Inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on an output of the department classification network model;

[0146] Matching the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient;

[0147] Based on the first matching degree corresponding to each candidate patient, a target patient matching the target patient is determined.

[0148] The computer storage medium of the embodiment of the present disclosure may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it.

[0149] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0150] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0151] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0152] It should be understood by those skilled in the art that the above modules or steps of the present disclosure can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present disclosure is not limited to any specific combination of hardware and software.

[0153] Note that the above are only preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art will understand that the present disclosure is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present disclosure. Therefore, although the present disclosure is described in more detail through the above embodiments, the present disclosure is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present disclosure, and the scope of the present disclosure is determined by the scope of the attached claims.

Claims

1. A medical triage method, It is characterized in that include: Obtain target chief complaint information of target patients; Inputting the target chief complaint information into a pre-trained emotion recognition network model, and determining the target emotion information of the target patient based on the output of the emotion recognition network model; Inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on an output of the department classification network model; Matching the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient; Based on the first matching degree corresponding to each candidate patient, a target patient matching the target patient is determined.

2. The method according to claim 1, It is characterized in that The emotion recognition network model includes: a first word vector generation sub-model, an emotion feature extraction sub-model and a first classification sub-model; Inputting the target complaint information into a pre-trained emotion recognition network model includes: Inputting the target chief complaint information into the first word vector generation sub-model for text conversion to generate a first word vector sequence; Inputting the first word vector sequence into the emotion feature extraction sub-model to extract context emotion features, and obtaining extracted target emotion feature information; The target emotion feature information is input into the first classification sub-model for emotion classification, and the predicted probability value corresponding to each emotion information is obtained and output.

3. The method according to claim 2, It is characterized in that The emotion feature extraction sub-model includes: a forward gated recurrent unit, a reverse gated recurrent unit and a feature concatenation unit; The step of inputting the first word vector sequence into the emotion feature extraction sub-model to extract context emotion features and obtain extracted target emotion feature information includes: Inputting the first word vector sequence into the forward gated recurrent unit to extract the sentiment features of the forward sequence, and obtaining the extracted first sentiment feature information; Inputting the first word vector sequence into the reverse gated recurrent unit to extract the emotion features of the reverse sequence, and obtaining the extracted second emotion feature information; The first emotion feature information and the second emotion feature information are input into the feature splicing unit for splicing to obtain target emotion feature information.

4. The method according to claim 1, It is characterized in that The department classification network model includes: at least one feature extraction sub-model and a second classification sub-model; Inputting the target chief complaint information into the pre-trained department classification network model includes: Inputting the target chief complaint information into each of the feature extraction sub-models to extract corresponding dimensional features, thereby obtaining extracted chief complaint feature information; The chief complaint characteristic information is input into the second classification sub-model for department classification, and the predicted probability value corresponding to each department is obtained and output.

5. The method according to claim 4, It is characterized in that The feature extraction sub-model includes: at least one of a word frequency-inverse text frequency statistics sub-model, a second word vector generation sub-model, a document topic generation sub-model and a part-of-speech extraction sub-model.

6. The method according to claim 1, It is characterized in that The method further comprises: Based on the historical conversation information corresponding to each patient in each department, determining the valid response information corresponding to each patient in each department; The effective reply information is input into a pre-trained personality recognition network model, and based on the output of the personality recognition network model, the personality information corresponding to each patient in each department is determined.

7. The method according to claim 6, It is characterized in that The determining of the valid response information corresponding to each patient in each department based on the historical conversation information corresponding to each patient in each department includes: Extracting historical response information from historical conversation information corresponding to each patient in each department; The drug names in the historical reply information are deleted, and the historical reply information after deletion whose length is greater than a preset length is used as valid reply information.

8. The method according to any one of claims 1 to 7, It is characterized in that The step of determining a target patient matching the target patient based on the first matching degree corresponding to each candidate patient includes: Obtaining first parameter information corresponding to the target patient and second parameter information corresponding to each candidate patient under preset matching parameters; Based on the first parameter information and the second parameter information, determining a second matching degree between each candidate patient and the target patient under preset matching parameters; Based on the first matching degree and the second matching degree corresponding to each candidate patient, a target patient matching the target patient is determined.

9. A medical triage device, It is characterized in that include: A target chief complaint information acquisition module is used to acquire the target chief complaint information of the target patient; A target emotion information determination module, used to input the target chief complaint information into a pre-trained emotion recognition network model, and determine the target emotion information of the target patient based on the output of the emotion recognition network model; A target department determination module, used for inputting the target chief complaint information into a pre-trained department classification network model, and determining a target department matching the target patient based on the output of the department classification network model; A first matching degree determination module is used to match the personality information of each candidate patient in the target department with the target emotion information to determine a first matching degree between each candidate patient and the target patient; The target patient determination module is used to determine a target patient matching the target patient based on the first matching degree corresponding to each candidate patient.

10. An electronic device, It is characterized in that The electronic device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the medical triage method as described in any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the medical triage method as described in any one of claims 1-8 is implemented.