Method and device for processing medical text information

Abstract

The present invention discloses a method and device for processing medical text information, which relates to the field of data processing technology and is applicable to the field of medical health. The main purpose is to solve the problem of low processing efficiency after the existing medical text information is classified. It includes: obtaining medical text information to be processed, the medical text information is input based on an electronic triage interactive platform; classifying the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, the text classification model is determined during the model training process based on the uncertainty score of the model classification layer and the patience coefficient to exit the model training; parsing the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, pushing the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform, so as to access the target electronic clinic in the electronic triage interactive platform.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, applicable to the field of medical health, and in particular to a method and device for processing medical text information. Background Art

[0002] With the rapid development of digital medical technology, electronic triage has become a common method for patients to seek medical treatment online. When patients enter electronic triage, medical text information is parsed to determine the specific electronic department, which is then pushed to the online doctor or electronic doctor for treatment.

[0003] At present, when processing medical text information, classification is usually performed directly based on text classification models. However, commonly used text classification models require a large amount of medical training data for training, which greatly wastes computing resources and increases the difficulty of application between actual medical text information. It is easy to fall into a model training deadlock, thereby greatly reducing the effectiveness of processing after medical text information classification. Summary of the Invention

[0004] In view of this, the present invention provides a method and apparatus for processing medical text information, the main purpose of which is to solve the problem of low processing efficiency of existing medical text information after classification.

[0005] According to one aspect of the present invention, a method for processing medical text information is provided, comprising:

[0006] Acquiring medical text information to be processed, wherein the medical text information is input based on the electronic triage interactive platform;

[0007] Classify the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit the model training;

[0008] Analyze the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform to access the target electronic clinic in the electronic triage interaction platform.

[0009] Furthermore, before the text classification model based on the completed model training performs classification processing on the medical text information and obtains the classification result of the medical text information, the method further includes:

[0010] Obtaining a text training sample set and constructing an initial text classification model, wherein the initial text classification model includes multiple model classification layers;

[0011] During model training of the initial text classification model based on the validation set of the text training sample set, determining the uncertainty score and the patience coefficient according to the classification results in the model classification layer, wherein the uncertainty score is used to characterize the probability of use of the connection layer, and the patience coefficient is used to characterize the situation of determining whether to exit model training based on the use probability;

[0012] If the uncertainty score and the patience coefficient reach a preset training exit condition, a target connection layer is determined, and model training is performed on the initial text classification model including the target connection layer based on the training set of the text training sample set.

[0013] Furthermore, determining the uncertainty score and the patience coefficient according to the classification results in the model classification layer includes:

[0014] determining a probability distribution of a classification layer of the model based on the classification results, and determining an uncertainty score based on an entropy of the probability distribution;

[0015] When the uncertainty score is less than a preset threshold, the patience coefficient is assigned 1 to obtain an updated patience coefficient.

[0016] Furthermore, the entity intent corresponding to the analysis of the classification result includes:

[0017] Retrieving an intention classification database, wherein the intention classification database stores words of entity classifications corresponding to different medical intentions;

[0018] If the words in the medical text information match the words of the entity classification in the intention classification database, the medical intention is determined to be the entity intention corresponding to the classification result.

[0019] Furthermore, the method further comprises:

[0020] Retrieving the index text of the electronic triage and screening the clinic features in the index text one by one, wherein the index text records the feature content configured according to the department level;

[0021] If the entity intends to match the clinic feature, it is determined as the matching target electronic clinic feature.

[0022] Furthermore, the access interface corresponding to the target electronic clinic feature is pushed to the electronic triage interactive platform, including:

[0023] Retrieving an access interface that matches the target electronic clinic characteristics from an interface library, wherein the interface library stores at least one access interface that matches different clinic characteristics;

[0024] If there are multiple access interfaces, an interface display dialog box is generated, and the interface display dialog box is displayed in the electronic triage interactive platform, so that the patient can select an access interface through the interface display dialog box;

[0025] If there is only one access interface, the access interface is pushed to the electronic triage interactive platform.

[0026] Furthermore, before obtaining the medical text information to be processed, the method further includes:

[0027] When it is detected that there are medical triage characteristic words in the medical consultation voice information, the medical consultation voice information is semantically converted to obtain the medical text information;

[0028] The medical text information is screened according to word entities in a medical word database to obtain medical text information to be processed, wherein the medical word database stores word entities associated with medical triage business requirements.

[0029] According to another aspect of the present invention, a device for processing medical text information is provided, comprising:

[0030] An acquisition module, configured to acquire medical text information to be processed, wherein the medical text information is inputted into the electronic triage interactive platform;

[0031] a processing module, configured to classify the medical text information based on a text classification model that has completed model training, to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit model training;

[0032] The push module is used to parse the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform, so as to access the target electronic clinic in the electronic triage interaction platform.

[0033] Furthermore, the device further comprises:

[0034] A construction module is used to obtain a text training sample set and construct an initial text classification model, wherein the initial text classification model includes multiple model classification layers;

[0035] a determination module configured to determine, during a model training process of the initial text classification model based on a validation set of the text training sample set, the uncertainty score and the patience coefficient according to a classification result in the model classification layer, wherein the uncertainty score is used to characterize the probability of use of the connection layer, and the patience coefficient is used to characterize whether to exit the model training by using the probability;

[0036] A training module is configured to determine a target connection layer if the uncertainty score and the patience coefficient reach a preset training exit condition, and perform model training on the initial text classification model including the target connection layer based on a training set of the text training sample set.

[0037] Furthermore, the determination module is specifically used to determine the probability distribution of the model classification layer based on the classification result, and determine the uncertainty score based on the entropy of the probability distribution; when the uncertainty score is less than a preset threshold, the patience coefficient is assigned 1 to obtain the updated patience coefficient.

[0038] Furthermore, the push module is specifically used to call the intention classification database, which stores words of entity classification corresponding to different medical intentions; if the words in the medical text information match the words of entity classification in the intention classification database, the medical intention is determined as the entity intention corresponding to the classification result.

[0039] Furthermore,

[0040] The retrieval module is also used to retrieve the index text of the electronic triage and screen the clinic features in the index text one by one. The index text records the feature content configured according to the department level; if the entity intends to match the clinic feature, it is determined as the matching target electronic clinic feature.

[0041] Furthermore, the push module is specifically used to retrieve an access interface that matches the characteristics of the target electronic clinic from an interface library, wherein the interface library stores at least one access interface that matches the characteristics of different clinics; if there are multiple access interfaces, an interface display dialog box is generated, and the interface display dialog box is displayed in the electronic triage interactive platform for the patient to select an access interface through the interface display dialog box; if there is only one access interface, the access interface is pushed to the electronic triage interactive platform.

[0042] Furthermore, the device further comprises:

[0043] a conversion module, configured to, when detecting the presence of medical triage characteristic words in the medical consultation voice information, perform semantic conversion on the medical consultation voice information to obtain the medical text information;

[0044] The screening module is used to screen the medical text information according to the word entities in the medical word database to obtain the medical text information to be processed. The medical word database stores word entities associated with medical triage business needs.

[0045] According to another aspect of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to execute an operation corresponding to the above-mentioned method for processing medical text information.

[0046] According to another aspect of the present invention, there is provided a computer device comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0047] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned method for processing medical text information.

[0048] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0049] The present invention provides a method and device for processing medical text information. Compared with the prior art, the embodiment of the present invention obtains medical text information to be processed, wherein the medical text information is input based on an electronic triage interactive platform; classifies the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on the uncertainty score of the model classification layer and the patience coefficient to exit the model training; analyzes the entity intention corresponding to the classification result, and when the entity intention matches the target electronic clinic feature, pushes the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform, so as to access the target electronic clinic in the electronic triage interactive platform, thereby optimizing model training based on early model exit, improving the accuracy of text classification using the model, reducing the computing resources used by the model, breaking out of the model training deadlock, and improving the accuracy of processing medical text information, thereby achieving the purpose of electronic triage based on the processing of medical text information.

[0050] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0052] Figure 1 A flow chart of a method for processing medical text information provided by an embodiment of the present invention is shown;

[0053] Figure 2 A block diagram showing the composition of a medical text information processing device provided by an embodiment of the present invention is shown;

[0054] Figure 3 A schematic structural diagram of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0055] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0056] Embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0057] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0058] Based on this, in one embodiment, an embodiment of the present invention provides a method for processing medical text information, which is illustrated by applying the method to a computer device such as a server. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, such as smart medical systems, digital medical platforms, etc.

[0059] The embodiment of the present invention provides a method for processing medical text information, such as Figure 1 As shown, the method includes:

[0060] 101. Obtain medical text information to be processed.

[0061] In an embodiment of the present invention, the current execution end, as the execution subject of the digital medical platform, can be a cloud server or a terminal server, that is, the current execution end can access outbound call requests from different patient users through the digital medical platform, thereby obtaining medical text information to be processed. In a specific implementation scenario, the user's outbound call request can be accessed through the electronic triage interaction platform in the digital medical platform. At this time, the medical text information is based on the input in the electronic triage interaction platform. For example, the patient enters the electronic triage interaction platform through the digital medical platform. At this time, the electronic triage interaction platform can instruct the patient to enter the medical text information through pre-configured text questions or voice questions. For example, if the patient enters "I want to consult about triage", the electronic triage interaction platform will output a voice prompt message such as "Please tell me your symptoms" or "Where do you feel uncomfortable?", thereby instructing the patient to enter the corresponding voice message "headache", and then obtain the text message "headache". The embodiment of the present invention does not make specific limitations.

[0062] It should be noted that since the medical text information to be processed is voice content or text content entered by the user in real time, when the current execution end executes step 101, it is executed after the voice content or text content is obtained in the electronic triage interactive platform. The embodiment of the present invention does not specifically limit the detection method of voice content or text content.

[0063] 102. Classify the medical text information based on the text classification model that has completed model training to obtain a classification result of the medical text information.

[0064] In an embodiment of the present invention, after receiving the medical text information to be processed, the current execution terminal classifies the medical text information based on a text classification model to obtain a classification result. In this case, the text classification model determines whether to exit model training during the model training process based on the uncertainty score and patience coefficient of the model classification layer. That is, during the text classification model training process, model completion is determined not based on model accuracy, but rather based on the uncertainty score and patience coefficient of the model classification layer. The uncertainty score is used to measure whether the probability distribution of the model classification layer is concentrated on a fixed category and can be calculated and determined using the probability distribution and its entropy. The patience coefficient is used to measure whether the uncertainty score of the model classification layer meets the model iteration requirements and can be determined by accumulating the number of uncertainty scores that match a preset score threshold. Model training is then stopped when the uncertainty score and patience coefficient match the conditions, thereby improving the efficiency and accuracy of model training. Furthermore, in an embodiment of the present invention, the classification results may include, but are not limited to, one or more medical department categories, such as neurology, cardiology, surgery, or orthopedics, etc., although this embodiment is not specifically limited thereto.

[0065] It should be noted that in the embodiments of the present invention, the initial text classification model used to construct the text classification model to be trained can be a pre-trained model for natural language task processing, including but not limited to the BERT model, and is not specifically limited in the embodiments of the present invention. Furthermore, the initial text classification model includes multiple model classification layers, so that uncertainty scores and patience coefficients can be calculated for each model classification layer during training.

[0066] 103. Analyze the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform.

[0067] In an embodiment of the present invention, after the current execution end determines the classification result, since the classification result may include multiple classification results, the classification result will be analyzed for entity intent, that is, the specific department expected to be triaged in the classification result will be determined. Among them, the entity is the word or text that distinguishes the departments from each other, and the entity intent is the entity that the patient hopes for or targets. At this time, the current execution end obtains it by parsing the classification result. After determining the entity intent, in order to accurately determine the clinic interface to be pushed to the patient, the current execution end matches the entity intent with the target electronic clinic feature. At this time, the target electronic clinic feature is the feature that marks each electronic clinic in the digital medical platform. For the electronic clinic pediatric orthopedic clinic 1, the corresponding clinic features are pediatrics and orthopedics, so that the expected electronic clinic is matched based on the entity intent. After determining that the target electronic clinic feature matches the entity intent, it means that the corresponding electronic clinic can be accessed. At this time, the current execution end pushes the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform to access the target electronic clinic in the electronic triage interactive platform.

[0068] It should be noted that, since the same clinic characteristics can correspond to multiple clinics, for example, Cardiology Clinic 1 and Cardiology Clinic 2, when pushing the access interface corresponding to the target electronic clinic characteristics, multiple access interfaces corresponding to all target electronic clinic characteristics can be pushed to the electronic triage interactive platform so that patients can choose by themselves. In addition, the electronic clinic is an interactive conversation environment built for doctors and patients through the digital medical platform. Doctors can see patients through the electronic clinic, and patients can see patients through the electronic clinic. At this time, doctors can see multiple patients at the same time, and the clinics between each patient are isolated from each other. At this time, the reception status of each target electronic clinic can also be obtained and counted at the same time. For example, if a doctor has seen 3 patients through the electronic clinic, when pushing the access interface, the information that 3 patients are being seen will be displayed in the electronic triage interactive platform so that patients can choose flexibly.

[0069] In another embodiment of the present invention, for further explanation and limitation, before the step of classifying the medical text information based on the text classification model for which model training has been completed and obtaining the classification result of the medical text information, the method further includes:

[0070] Obtain a text training sample set and build an initial text classification model;

[0071] In a process of training the initial text classification model based on a validation set of the text training sample set, determining the uncertainty score and the patience coefficient according to classification results in the model classification layer;

[0072] If the uncertainty score and the patience coefficient reach a preset training exit condition, a target connection layer is determined, and model training is performed on the initial text classification model including the target connection layer based on the training set of the text training sample set.

[0073] To achieve the purpose of classifying medical text information based on a text classification model, thereby improving the processing efficiency of medical text information, the current execution end pre-acquires a text training sample set. In this case, the text training sample set includes classification labels marked by different text samples. The classification labels can be obtained by manually marking historical text, including but not limited to the same medical department classification as the classification result, which is not specifically limited in this embodiment of the present invention. At the same time, the current execution end can also construct an initial text classification model, that is, the initial text classification model includes multiple model classification layers. If the initial text classification model is a natural language pre-trained model, the corresponding model classification layer is the classification layer connected to the Transformer layer after the Embedding layer. That is, the initial text classification model includes multiple classification layers connected in sequence to the Transformer layer and correspondingly connected to the Transformer layer, so as to calculate the uncertainty score and patience coefficient of each Transformer layer during the training process. The current execution end selects a specific number of samples from the text training sample set as a validation set and trains the initial text classification model based on the validation set. During the training process, the uncertainty score and patience coefficient are calculated based on the classification results of each model classification layer.

[0074] It should be noted that when it is determined that the model training is completed, the target connection layer is determined, that is, the Transformer layer corresponding to the uncertainty score less than the preset iteration requirement threshold is selected as the target connection layer, so that the initial text classification model including the target connection layer is trained based on the training set of the text training sample set. The embodiment of the present invention does not specifically limit the screening of the training set.

[0075] In another embodiment of the present invention, for further illustration and limitation, the step of determining the uncertainty score and the patience coefficient based on the classification results in the model classification layer includes:

[0076] determining a probability distribution of a classification layer of the model based on the classification results, and determining an uncertainty score based on an entropy of the probability distribution;

[0077] When the uncertainty score is less than a preset threshold, the patience coefficient is assigned 1 to obtain an updated patience coefficient.

[0078] In order to accurately determine the uncertainty score and patience coefficient, thereby improving the efficiency and flexibility of model training, specifically, since the uncertainty score is used to characterize whether the probability distribution of the model classification layer is concentrated on a fixed classification, it can be calculated and determined using the probability distribution and the entropy of the probability distribution, that is, the uncertainty score calculation method for a certain layer of the model classification layer is: first, based on the classification results in the model training process (such as the number of training samples of the internal medicine classification, the number of training samples of the surgical classification), calculate the probability distribution belonging to this model classification layer; calculate the entropy of this probability distribution, and combine the maximum probability value in the probability distribution to determine the uncertainty score of this model classification (in this process, the uncertainty scores corresponding to different entropy and maximum probability values ​​can be pre-configured). Since the patience coefficient is used to characterize whether the uncertainty score of the model classification layer meets the model iteration requirements, it can be determined by accumulating the number of times the uncertainty score matches the preset score threshold. That is, the patience coefficient calculation method for a certain layer of the model classification layer is: after calculating the uncertainty score, it is determined that the uncertainty score is less than the preset iteration requirement threshold, and the patience coefficient is added by 1, thereby superimposing; when the uncertainty score is greater than or equal to the preset iteration requirement threshold, the patience coefficient is reset to 0, until the patience coefficient is greater than the preset patience value, it is determined that the model training process of the initial text model is completed, and the purpose of determining the structure of the model is achieved. That is, in the process of building the model, after screening the classification layer, the screening of the model structure based on the training process is exited, and model training is continued. At this time, the preset training exit condition is that the total value of the patience coefficient after superposition is greater than the preset patience value. The embodiment of the present invention does not specifically limit the preset patience value.

[0079] In another embodiment of the present invention, for further explanation and limitation, the step of parsing the entity intent corresponding to the classification result includes:

[0080] Retrieve the intent classification database;

[0081] If the words in the medical text information match the words of the entity classification in the intention classification database, the medical intention is determined to be the entity intention corresponding to the classification result.

[0082] In order to determine the intent in the classification results and match it to the accurate electronic clinic, thereby improving the processing efficiency of medical text information, the current execution end first retrieves the intent classification database to match the words in the text information based on the intent classification database to determine the entity intent. Among them, the intent classification database stores words of entity classifications corresponding to different medical intentions. Developers can enter and configure when creating the intent classification database, and update it at specific time intervals, so as to achieve the purpose of flexible and accurate identification of intent classification. After the current execution end retrieves the intent classification database, it compares each word in the medical text information with the words corresponding to the entity classification in the intent classification database. If the words in the medical text information match the words corresponding to the entity classification, it means that the medical text information has a medical intent corresponding to the entity classification. Therefore, this medical intent can be driven into the entity intent of the classification result. The embodiment of the present invention does not make specific limitations.

[0083] In another embodiment of the present invention, for further explanation and limitation, the steps further include:

[0084] Retrieving the index text of the electronic triage and screening the clinic features in the index text one by one;

[0085] If the entity intends to match the clinic feature, it is determined as the matching target electronic clinic feature.

[0086] In order to accurately identify the electronic clinic that the patient is expected to access, thereby improving the efficiency and effectiveness of processing medical text information, the current execution end filters the electronic clinics by retrieving the index text of the electronic triage. Among them, the index text records the feature content configured according to the department classification. The department classification is used to characterize the classification between the major departments and the minor departments. For example, the major departments are internal medicine, surgery, pediatrics, etc., and the minor departments are cardiology, orthopedics, etc. The embodiment of the present invention does not make specific limitations. In the embodiment of the present invention, when the current execution end configures each feature content, that is, it records the feature words corresponding to each department classification in the index text as clinic features, including but not limited to "child", "bone", "heart", "stomach", etc., so as to match the entity intent with the clinic features of each electronic clinic. When the entity intent matches the clinic features of the electronic clinic, the electronic clinic is determined as the target electronic clinic, and then the target electronic clinic features are determined. At this time, the target electronic clinic features are the clinic features that match the entity intent. The embodiment of the present invention does not make specific limitations.

[0087] In another embodiment of the present invention, for further explanation and limitation, the step of pushing the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform includes:

[0088] Retrieving an access interface that matches the characteristics of the target electronic clinic from an interface library;

[0089] If there are multiple access interfaces, an interface display dialog box is generated, and the interface display dialog box is displayed in the electronic triage interactive platform, so that the patient can select an access interface through the interface display dialog box;

[0090] If there is only one access interface, the access interface is pushed to the electronic triage interactive platform.

[0091] To push access interfaces for different electronic clinics to patients, thereby improving the efficiency of processing medical text information, the current execution end, when pushing an access interface, first retrieves an access interface that matches the characteristics of the target electronic clinic from an interface library and pushes this access interface to the electronic triage interactive platform. The interface library stores at least one access interface that matches the characteristics of different clinics, so that the access interface corresponding to the characteristics of the target electronic clinic is sequentially matched from the interface library and pushed to the electronic triage interactive platform for patients to select and access.

[0092] It should be noted that, since multiple entity intentions can be identified in medical text information, and then multiple clinic features and multiple corresponding access interfaces can be matched, in a specific implementation scenario, if there are multiple access interfaces, the current execution end generates an interface display dialog box, and displays the interface display dialog box in the electronic triage interactive platform, so that the patient can select the access interface through the interface display dialog box. If there is only one access interface, this access interface can be directly pushed to the electronic triage interactive platform. In addition, the access interface in the embodiment of the present invention is the link address to enter the target electronic clinic, so that the patient can jump to the conversation box of the target electronic clinic in the electronic triage interactive platform to interact with the doctor through the digital medical platform. In the embodiment of the present invention, there is no specific limitation on the form of the conversation box.

[0093] In another embodiment of the present invention, for further explanation and limitation, before the step of obtaining the medical text information to be processed, the method further includes:

[0094] When it is detected that there are medical triage characteristic words in the medical consultation voice information, the medical consultation voice information is semantically converted to obtain the medical text information;

[0095] The medical text information is screened according to word entities in a medical word database to obtain medical text information to be processed, wherein the medical word database stores word entities associated with medical triage business requirements.

[0096] In order to achieve the purpose of effectively processing medical text information, and thus realize electronic triage based on the processing of medical text information, when the current execution end obtains medical text information through the electronic triage interactive platform, it can specifically detect whether there are medical triage feature words in the patient's medical consultation voice information. Among them, the medical triage feature words are used to indicate the words that the current execution end determines whether triage is needed, such as the words "I have a headache, which department should I see?" In this case, the medical triage feature words can be in voice form, and the embodiment of the present invention does not make specific limitations. When medical triage feature words exist in the medical consultation voice information, it means that triage is needed. Therefore, the current execution end performs semantic conversion on the medical consultation voice information to obtain all the text content, so as to filter the medical text information according to the word entities in the medical word database to obtain the medical text information to be processed. Among them, the medical word database stores word entities related to medical triage business needs, that is, other words in "I feel unwell, nauseous, and have a headache. Which department should I go to?" are deleted, and only words such as "headache", "nausea", and "body" corresponding to the medical triage business needs are retained as medical text information to be processed. The embodiment of the present invention does not make specific limitations.

[0097] An embodiment of the present invention provides a method for processing medical text information. Compared with the prior art, the embodiment of the present invention obtains medical text information to be processed, wherein the medical text information is entered in an electronic triage interactive platform; classifies the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on the uncertainty score of the model classification layer and the patience coefficient to exit the model training; analyzes the entity intention corresponding to the classification result, and when the entity intention matches the target electronic clinic feature, pushes the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform, so as to access the target electronic clinic in the electronic triage interactive platform, thereby optimizing model training based on early model exit, improving the accuracy of text classification using the model, reducing the computing resources used by the model, breaking out of the model training deadlock, and improving the accuracy of processing medical text information, thereby achieving the purpose of electronic triage based on the processing of medical text information.

[0098] Furthermore, as a response to the above Figure 1 The embodiment of the present invention provides a device for processing medical text information, such as Figure 2 As shown, the device includes:

[0099] An acquisition module 21 is used to acquire medical text information to be processed, where the medical text information is input into the electronic triage interactive platform;

[0100] a processing module 22 configured to classify the medical text information based on a text classification model that has completed model training, to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit model training;

[0101] The push module 23 is used to parse the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform to access the target electronic clinic in the electronic triage interaction platform.

[0102] Furthermore, the device further comprises:

[0103] A construction module is used to obtain a text training sample set and construct an initial text classification model, wherein the initial text classification model includes multiple model classification layers;

[0104] a determination module configured to determine, during a model training process of the initial text classification model based on a validation set of the text training sample set, the uncertainty score and the patience coefficient according to a classification result in the model classification layer, wherein the uncertainty score is used to characterize the probability of use of the connection layer, and the patience coefficient is used to characterize whether to exit the model training by using the probability;

[0105] A training module is configured to determine a target connection layer if the uncertainty score and the patience coefficient reach a preset training exit condition, and perform model training on the initial text classification model including the target connection layer based on a training set of the text training sample set.

[0106] Furthermore, the determination module is specifically used to determine the probability distribution of the model classification layer based on the classification result, and determine the uncertainty score based on the entropy of the probability distribution; when the uncertainty score is less than a preset threshold, the patience coefficient is assigned 1 to obtain the updated patience coefficient.

[0107] Furthermore, the push module is specifically used to call the intention classification database, which stores words of entity classification corresponding to different medical intentions; if the words in the medical text information match the words of entity classification in the intention classification database, the medical intention is determined as the entity intention corresponding to the classification result.

[0108] Furthermore,

[0109] The retrieval module is also used to retrieve the index text of the electronic triage and screen the clinic features in the index text one by one. The index text records the feature content configured according to the department level; if the entity intends to match the clinic feature, it is determined as the matching target electronic clinic feature.

[0110] Furthermore, the push module is specifically used to retrieve an access interface that matches the characteristics of the target electronic clinic from an interface library, wherein the interface library stores at least one access interface that matches the characteristics of different clinics; if there are multiple access interfaces, an interface display dialog box is generated, and the interface display dialog box is displayed in the electronic triage interactive platform for the patient to select an access interface through the interface display dialog box; if there is only one access interface, the access interface is pushed to the electronic triage interactive platform.

[0111] Furthermore, the device further comprises:

[0112] a conversion module, configured to, when detecting the presence of medical triage characteristic words in the medical consultation voice information, perform semantic conversion on the medical consultation voice information to obtain the medical text information;

[0113] The screening module is used to screen the medical text information according to the word entities in the medical word database to obtain the medical text information to be processed. The medical word database stores word entities associated with medical triage business needs.

[0114] An embodiment of the present invention provides a device for processing medical text information. Compared with the prior art, the embodiment of the present invention obtains medical text information to be processed, wherein the medical text information is entered in an electronic triage interactive platform; classifies the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on the uncertainty score of the model classification layer and the patience coefficient to exit the model training; analyzes the entity intention corresponding to the classification result, and when the entity intention matches the target electronic clinic feature, pushes the access interface corresponding to the target electronic clinic feature to the electronic triage interactive platform, so as to access the target electronic clinic in the electronic triage interactive platform, thereby optimizing model training based on early model exit, improving the accuracy of text classification using the model, reducing the computing resources used by the model, breaking out of the model training deadlock, and improving the accuracy of processing medical text information, thereby achieving the purpose of electronic triage based on the processing of medical text information.

[0115] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction. The computer-executable instruction can execute the method for processing medical text information in any of the above method embodiments.

[0116] Figure 3 A schematic structural diagram of a computer device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

[0117] like Figure 3 As shown, the computer device may include: a processor 302 , a communication interface 304 , a memory 306 , and a communication bus 308 .

[0118] The processor 302 , the communication interface 304 , and the memory 306 communicate with each other via a communication bus 308 .

[0119] The communication interface 304 is used to communicate with other devices such as clients or other servers.

[0120] The processor 302 is configured to execute the program 310 , and specifically to execute the relevant steps in the above-mentioned embodiment of the method for processing medical text information.

[0121] Specifically, the program 310 may include program codes, which include computer operation instructions.

[0122] Processor 302 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in a computer device may be of the same type, such as one or more CPUs, or may be of different types, such as one or more CPUs and one or more ASICs.

[0123] The memory 306 is used to store the program 310. The memory 306 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0124] The program 310 may be specifically configured to cause the processor 302 to perform the following operations:

[0125] Acquiring medical text information to be processed, wherein the medical text information is input based on the electronic triage interactive platform;

[0126] Classify the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit the model training;

[0127] Analyze the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform to access the target electronic clinic in the electronic triage interaction platform.

[0128] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, centralized on a single computing device, or distributed across a network of multiple computing devices. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. In some cases, the steps shown or described can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0129] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for processing medical text information, characterized in that: include: Acquiring medical text information to be processed, wherein the medical text information is input based on the electronic triage interactive platform; Classify the medical text information based on a text classification model that has completed model training to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit the model training; parsing the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, pushing the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform, so as to access the target electronic clinic in the electronic triage interaction platform; Before classifying the medical text information based on the text classification model that has completed model training to obtain the classification result of the medical text information, the method further includes: Obtaining a text training sample set and constructing an initial text classification model, wherein the initial text classification model includes multiple model classification layers; During model training of the initial text classification model based on the validation set of the text training sample set, determining the uncertainty score and the patience coefficient according to the classification results in the model classification layer, wherein the uncertainty score is used to characterize the probability of use of the connection layer, and the patience coefficient is used to characterize the situation of determining whether to exit model training based on the use probability; If the uncertainty score and the patience coefficient reach a preset training exit condition, determining a target connection layer, and performing model training on the initial text classification model including the target connection layer based on the training set of the text training sample set; Determining the uncertainty score and the patience coefficient according to the classification results in the model classification layer includes: determining a probability distribution of a classification layer of the model based on the classification results, and determining an uncertainty score based on an entropy of the probability distribution; When the uncertainty score is less than a preset threshold, the patience coefficient is assigned 1 to obtain an updated patience coefficient.

2. The method according to claim 1, characterized in that The entity intent corresponding to the analysis of the classification result includes: Retrieving an intention classification database, wherein the intention classification database stores words of entity classifications corresponding to different medical intentions; If the words in the medical text information match the words of the entity classification in the intention classification database, the medical intention is determined to be the entity intention corresponding to the classification result.

3. The method according to claim 2, characterized in that The method further comprises: Retrieving the index text of electronic triage and screening the clinic features in the index text one by one, wherein the index text records the feature content configured according to the department level; If the entity intends to match the clinic feature, it is determined as the matching target electronic clinic feature.

4. The method according to claim 1, wherein The access interface corresponding to the target electronic clinic feature is pushed to the electronic triage interactive platform, including: Retrieving an access interface that matches the target electronic clinic characteristics from an interface library, wherein the interface library stores at least one access interface that matches different clinic characteristics; If there are multiple access interfaces, an interface display dialog box is generated, and the interface display dialog box is displayed in the electronic triage interactive platform, so that the patient can select an access interface through the interface display dialog box; If there is only one access interface, the access interface is pushed to the electronic triage interactive platform.

5. The method according to any one of claims 1 to 4, characterized in that Before obtaining the medical text information to be processed, the method further includes: When it is detected that there are medical triage characteristic words in the medical consultation voice information, the medical consultation voice information is semantically converted to obtain the medical text information; The medical text information is screened according to word entities in a medical word database to obtain medical text information to be processed, wherein the medical word database stores word entities associated with medical triage business requirements.

6. A device for processing medical text information, characterized in that: include: An acquisition module, configured to acquire medical text information to be processed, wherein the medical text information is inputted into the electronic triage interactive platform; a processing module, configured to classify the medical text information based on a text classification model that has completed model training, to obtain a classification result of the medical text information, wherein the text classification model is determined during the model training process based on an uncertainty score of a model classification layer and a patience coefficient to exit model training; a push module, configured to analyze the entity intent corresponding to the classification result, and when the entity intent matches the target electronic clinic feature, push the access interface corresponding to the target electronic clinic feature to the electronic triage interaction platform, so as to access the target electronic clinic in the electronic triage interaction platform; The device further comprises: A construction module is used to obtain a text training sample set and construct an initial text classification model, wherein the initial text classification model includes multiple model classification layers; a determination module configured to determine, during a model training process of the initial text classification model based on a validation set of the text training sample set, the uncertainty score and the patience coefficient according to a classification result in the model classification layer, wherein the uncertainty score is used to characterize the probability of use of the connection layer, and the patience coefficient is used to characterize whether to exit the model training by using the probability; a training module, configured to determine a target connection layer if the uncertainty score and the patience coefficient meet a preset training exit condition, and perform model training on the initial text classification model including the target connection layer based on a training set of the text training sample set; The determination module is specifically configured to determine a probability distribution of the model classification layer based on the classification result, and to determine an uncertainty score based on the entropy of the probability distribution; when the uncertainty score is less than a preset threshold, assign 1 to the patience coefficient to obtain an updated patience coefficient.

7. A storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to execute an operation corresponding to the method for processing medical text information according to any one of claims 1 to 5.

8. A computer device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute an operation corresponding to the method for processing medical text information according to any one of claims 1 to 5.

Images