Rescue guidance method and device and medium

Abstract

The invention discloses a rescue guidance method and device and a medium, and the method comprises the steps: obtaining the help data of a help seeker; based on the help seeking data, determining the severity of a help seeking event corresponding to the help seeker; based on the severity, determining a target rescue guidance strategy for the help seeker; and based on the target rescue guidance strategy, guiding the help seeker to process the help event.

Description

Technical Field

[0001] The present application relates to the field of rescue guidance technology, and in particular to a rescue guidance method, device and medium. Background Art

[0002] In the related art, a pre-trained dispatcher usually provides remote rescue guidance to the help-seeker based on the help-seeking information provided by the help-seeker. However, such rescue guidance method, on the one hand, relies heavily on the rescue knowledge of the dispatcher, and puts forward higher requirements on the professional level of the dispatcher. On the other hand, the above rescue guidance method cannot meet the rescue guidance needs of a large number of help-seekers at the same time. Summary of the invention

[0003] Based on the above problems, the embodiments of the present application provide a rescue guidance method, device and medium.

[0004] The technical solution provided by the embodiment of the present application is as follows:

[0005] The present application provides a rescue guidance method, the method comprising:

[0006] Obtain the help-seeking data of the help-seeker;

[0007] Based on the help-seeking data, determining the severity of the help-seeking event corresponding to the help-seeker;

[0008] Based on the severity, determining a target rescue guidance strategy for the help-seeker;

[0009] Based on the target rescue guidance strategy, the help seeker is guided to handle the help-seeking event.

[0010] An embodiment of the present application also provides an electronic device, which includes a processor and a memory; a computer program is stored in the memory; when the computer program is executed by the processor, it can implement any of the rescue guidance methods described above.

[0011] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored; when the computer program is executed by a processor of an electronic device, it can implement any of the rescue guidance methods described above.

[0012] The rescue guidance method provided in the embodiment of the present application determines the severity of the help-seeking event corresponding to the help-seeker based on the help-seeking data after obtaining the help-seeking data of the help-seeker. Thus, through the above operation, a preliminary classification of the risk level of the help-seeking event corresponding to the help-seeker can be achieved; and, by determining the target rescue guidance strategy for the rescuer based on the severity, on the one hand, the target rescue guidance strategy is directly related to the actual severity of the help-seeker. Then, in the process of guiding the help-seeker to handle the help-seeking event through the target rescue guidance strategy, targeted and efficient rescue guidance for the help-seeker can be achieved. On the other hand, since the target rescue guidance strategy is targeted at the help-seeker, differentiated and targeted rescue guidance for the help-seeking events corresponding to different help-seekers of the same severity can be achieved. At the same time, the rescue guidance method provided in the embodiment of the present application can automatically and intelligently determine the target rescue guidance strategy for the rescuer, thereby reducing the dependence on the dispatcher and his professional and technical level. In the case of a large number of help-seekers, with the help of the powerful data processing capability and efficient data computing capability of the electronic equipment, parallel and efficient rescue guidance for a large number of help-seekers can be achieved, thereby meeting the actual rescue guidance needs for handling help-seeking events. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flowchart of a rescue guidance method provided in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of a process for obtaining help-seeking data provided in an embodiment of the present application;

[0015] Figure 3 A schematic diagram of the structure of the critical level classification model provided in the embodiment of the present application;

[0016] Figure 4 A schematic diagram of a process for determining a high-risk level provided in an embodiment of the present application;

[0017] Figure 5 A schematic diagram of the structure of the first model provided in the embodiment of the present application;

[0018] Figure 6 A schematic diagram of a process for determining a high-risk target rescue guidance strategy provided in an embodiment of the present application;

[0019] Figure 7 A schematic diagram of the structure of the first model in the initial state of training provided in an embodiment of the present application;

[0020] Figure 8 A schematic diagram of the process flow of the plan jump provided in the embodiment of the present application;

[0021] Fig. 9 Another schematic diagram of a flow chart of a plan jump provided in an embodiment of the present application;

[0022] Fig.10 A schematic diagram of the structure of the second model provided in the embodiment of the present application;

[0023] Fig.11 A schematic diagram of a process for determining a common patient guidance strategy provided in an embodiment of the present application;

[0024] Fig.12 A schematic diagram of a process for setting a level identification according to an embodiment of the present application;

[0025] Fig.13 A schematic diagram of a process for rescuing a person seeking help provided in an embodiment of the present application;

[0026] Fig.14 A schematic diagram of the structure of the rescue guidance system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0028] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] In emergency dispatch and remote rescue scenarios, the Medical Priority Dispatch System (MPDS) is a widely used rescue guidance system. In actual emergency dispatch and remote rescue guidance scenarios, dispatchers use the system to assess the condition, prioritize, dispatch emergency resources, and provide online guidance to help-seekers to save themselves or transfer them to ambulances in a timely manner.

[0030] In the above scenario, in order to improve the professional level of the dispatcher, the dispatcher needs to be fully trained in advance so that the dispatcher can conduct multiple rounds of inquiries to the help-seekers in the actual rescue scenario in combination with the prompts of the MPDS, and then provide the help-seekers with appropriate rescue guidance based on the answers provided by the help-seekers to the multiple rounds of inquiries.

[0031] However, the above rescue guidance process relies heavily on the dispatcher's knowledge of emergency dispatch, which places higher professional requirements on the dispatcher; moreover, the efficiency of multiple rounds of inquiries and answer collection between the dispatcher and the help seeker is low and cannot meet the help needs of a large number of help seekers.

[0032] Based on the above problems, the embodiments of the present application provide a rescue guidance method, device and medium.

[0033] The rescue guidance method provided in the embodiment of the present application can be implemented by a processor of an electronic device. The above-mentioned processor can be at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller (MCU), and a microprocessor.

[0034] Exemplarily, the electronic device may be a physical device or a virtual machine device.

[0035] Exemplarily, the electronic device may be a computer device or a robot.

[0036] Figure 1 A flowchart of the rescue guidance method provided in the embodiment of the present application is shown in FIG. Figure 1 As shown, the process may include the following steps:

[0037] Step 101: Obtain the help-seeking data of the help-seeker.

[0038] In one embodiment, the caller may include a person or animal that is injured or in distress.

[0039] In one embodiment, the help-seeking data may be used to characterize the current injury state, illness state, or predicament state of the help-seeker, and may also include the help-seeker's subjective feelings.

[0040] In one implementation, the help-seeking data may also include the environmental status of the environment where the help-seeker is located.

[0041] In one embodiment, the help-seeking data may include the chief complaint data of the person seeking help, and may also include standardized medical expression data obtained by processing the chief complaint data.

[0042] In one implementation, the help-seeking data may include at least one of audio data, video data, image data, geographic location data, and text data; illustratively, the geographic location data may indicate the location of the help-seeker.

[0043] In one embodiment, the person seeking help or a person associated with the person seeking help can establish a communication connection between the communication device and the electronic device by operating the communication device, and through the above communication connection, send relevant information such as the reason, status and purpose of the request for help to the electronic device by voice input, video acquisition, image acquisition, location sharing or text sharing, and the electronic device processes the above relevant information to obtain the help data.

[0044] In one implementation, a communication connection may be established with a smart device worn by the person seeking help, and the vital sign data of the person seeking help may be obtained through the communication connection, and then the vital sign data may be processed to obtain the help-seeking data.

[0045] Figure 2 A schematic diagram of the process of obtaining help-seeking data provided in an embodiment of the present application. Figure 2 As shown, the process may include the following steps:

[0046] Step A1: Obtain medical literature.

[0047] Step A2: Entity extraction.

[0048] For example, entities such as patient status, disease course, and drugs described in medical literature may be extracted to obtain entity extraction results.

[0049] Step A3: Relation extraction.

[0050] Exemplarily, the relationship between entities in the entity extraction result can be extracted according to the medical entity association rule to obtain the relationship extraction result.

[0051] Step A4: Form a knowledge graph.

[0052] Exemplarily, the above entity extraction results and relationship extraction results can be displayed in the form of a graph to form a knowledge graph.

[0053] Step A5: Obtain the medical condition text.

[0054] Exemplarily, after a call comes in from a person seeking help, voice data of the person seeking help may be recorded, and the main complaint content and / or vital signs data of the person seeking help may be extracted from the voice data, and then the main complaint content may be converted into text data to obtain a medical condition text; exemplary, the medical condition text may be unstructured text data.

[0055] Step A6: object extraction.

[0056] For example, since the person seeking help may be a non-professional medical staff, the main complaint may contain non-professional description of the condition. Therefore, data objects such as physiological state, psychological state, and medication records in the condition text can be extracted to obtain object extraction results.

[0057] Step A7: Entity alignment.

[0058] Exemplarily, the data objects can be mapped to entities based on the correspondence between the data objects in the object extraction results and the entities in the knowledge graph.

[0059] Step A8: entity correction.

[0060] Exemplarily, the association relationship between entities, the order in which entities appear, etc. may be corrected to obtain a correction result.

[0061] Step A9: Obtain help request data.

[0062] Exemplarily, the help-seeking data may be a correction result.

[0063] pass Figure 2 The process shown can realize the pre-processing of the chief complaint content and / or vital signs data provided by the help-seeker, so as to standardize, integrate and correct the discrete, one-sided and erroneous data in the chief complaint content through medical terminology, so that the help-seeking data obtained in the above manner can more comprehensively reflect the current actual status of the help-seeker.

[0064] Step 102: Determine the severity of the help-seeking event corresponding to the help-seeker based on the help-seeking data.

[0065] In one embodiment, the help-seeking event may be directly associated with the help-seeker; illustratively, the help-seeking event may include physical and / or psychological trauma events of the help-seeker himself, for example, trauma events may include fractures, chest tightness, etc.; illustratively, the help-seeking event may include an unexpected event in the help-seeker's environment where the threat level is greater than the threat threshold, for example, the unexpected event may include a natural disaster in the help-seeker's environment, such as a flood or fire.

[0066] In one implementation, the severity level may include the risk level of the current patient condition of the person seeking help; illustratively, the severity level may include high risk, normal risk, and the like.

[0067] In one implementation, the severity may include the threat level of the unexpected event occurring in the current environment of the help-seeker to the environment or the help-seeker.

[0068] In one embodiment, severity may be determined by any of the following:

[0069] A first correlation between the help-seeking data and the severity is obtained, and then the severity is determined based on the first correlation and the help-seeking data.

[0070] The help-seeking data is classified by the critical level classification model to determine the severity of the help-seeker's condition. Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of the critical level classification model provided in the embodiment of the present application. Figure 3 As shown, the critical level classification model 3 may include a text input module 301 , a text digitization module 302 , a deep learning module 303 and a vector output module 304 .

[0071] Among them, the content of the chief complaint and / or vital signs data can be Figure 2 The process shown is converted into text-type help data, which can be input into the text digitization module 302 through the text input module 301, so that the text digitization module 302 can segment the text data and perform numerical encoding on the segmentation results, thereby realizing text digitization processing.

[0072] Exemplarily, the deep learning module 303 may include a deep bidirectional language encoding representation (Bidirectional Encoder Representation from Transformers, BERT), which is used to extract deep features from the digitized results output by the text digitization module 302, and output the deep features through the vector output module 304; exemplarily, the output end of BERT can also be connected to the fully connected layer and the softmax layer, so as to output the probability of whether the person seeking help is at a high-risk level.

[0073] Exemplarily, the critical level classification model may be a binary classification model, through which the help-seeking data may be processed in a binary manner, thereby determining whether the help-seeker is at a high-risk level.

[0074] When the critical level classification model is a binary classification model, Figure 4 The process shown determines whether the severity of the person seeking help is at a high risk level. Figure 4 A schematic diagram of a process for determining a high-risk level provided in an embodiment of the present application, such as Figure 4 As shown, the process may include the following steps:

[0075] Step B1, obtaining the chief complaint data.

[0076] Exemplarily, the main complaint data may be the main complaint content in the aforementioned embodiments.

[0077] Step B2: pretreatment.

[0078] For example, it can be achieved by Figure 2 The process shown pre-processes the chief complaint data and / or vital sign data to obtain the help-seeking data.

[0079] Step B3: Binary classification processing.

[0080] Exemplarily, the help-seeking data may be processed by a binary classification model, wherein the binary classification model may be a critical level classification model.

[0081] Step B4: Determine whether it is high risk.

[0082] Exemplarily, whether the person seeking help is at a high-risk level can be determined through the output result of the binary classification model.

[0083] Table 1 is a statistical result of help-seeking data and critical level obtained by preprocessing different chief complaint data through the steps provided in the above embodiments provided in the embodiments of the present application. It can be seen from Table 1 that the help-seeking data obtained by preprocessing the chief complaint data can more accurately characterize the actual state of the help-seeker, and the critical level obtained by classifying the help-seeking data not only realizes the preliminary classification of the help-seeker, but also can more intuitively and vividly express the actual risk level of the help-seeker.

[0084] Step 103: Based on the severity, determine the target rescue guidance strategy for the help seeker.

[0085] Table 1

[0086]

[0087] In one embodiment, the target rescue guidance strategy may include the steps involved in providing rescue guidance to the person seeking help, as well as the jump conditions between the various steps, etc.; illustratively, the jump conditions may include the state of the person seeking help after the execution of the previous step is completed, whether the execution conditions of the next step are met, etc.

[0088] In one implementation, the target rescue guidance strategy may be determined by any of the following methods:

[0089] There may be a second correlation between the severity and the rescue guidance strategy. Thus, after the severity of the help-seeker is determined, a target rescue guidance strategy for the help-seeker may be determined based on the severity of the help-seeker and the second correlation.

[0090] The severity is identified through a neural network model to dynamically generate the steps and methods required to rescue the person seeking help, and the above steps and methods are determined as the target rescue guidance strategy.

[0091] Step 104: Based on the target rescue guidance strategy, guide the help-seeker to handle the help-seeking event.

[0092] In one embodiment, the steps and operational procedures in the target rescue guidance strategy can be output to the person seeking help or the target person to guide the person seeking help to rescue himself or herself and / or guide the target person to rescue the person seeking help, and can also guide the person seeking help and / or the target person to deal with unexpected events; illustratively, the target person may include a person who is associated with the person seeking help and has the ability to rescue.

[0093] As can be seen from the above, the rescue guidance method provided in the embodiment of the present application, after obtaining the help-seeking data of the help-seeker, determines the severity of the help-seeking event corresponding to the help-seeker based on the help-seeking data. In this way, through the above operation, it is possible to achieve a preliminary classification of the risk level of the help-seeking event corresponding to the help-seeker; and, by determining the target rescue guidance strategy for the rescuer based on the severity, on the one hand, the target rescue guidance strategy is directly related to the actual severity of the help-seeker, so in the process of guiding the help-seeker to handle the help-seeking event through the target rescue guidance strategy, it is possible to achieve targeted and efficient rescue guidance for the help-seeker, on the other hand, because the target rescue guidance strategy is targeted at the help-seeker, it is possible to achieve differentiated and targeted rescue guidance for the help-seeking events corresponding to different help-seekers of the same severity; at the same time, the rescue guidance method provided in the embodiment of the present application can automatically and intelligently determine the target rescue guidance strategy for the rescuer, thereby reducing the dependence on the dispatcher and his professional and technical level; in the case of a large number of help-seekers, with the help of the powerful data processing capability and efficient data computing capability of the electronic equipment, it is possible to achieve parallel and efficient rescue guidance for a large number of help-seekers, thereby meeting the actual rescue guidance needs for handling help-seeking events.

[0094] Based on the above embodiments, in the rescue guidance method provided in the embodiments of the present application, determining the target rescue guidance strategy for the help seeker based on the severity can be achieved by the following steps:

[0095] Step C1: Determine a target network model associated with the severity.

[0096] In one embodiment, a third association relationship may be established between the severity and the network model used to determine the rescue guidance category. Thus, after the severity is determined, the target network model may be determined based on the severity and the third association relationship.

[0097] Step C2: Obtain patient status data.

[0098] In one embodiment, the patient status data may include data such as at least one of the following states of the help-seeker: being sick, injured, and trapped, and the duration of the at least one state.

[0099] In one embodiment, the patient status data may include the historical life data and medical and medication data of the person seeking help; exemplarily, the historical life data may include the diet, exercise, rest, work and study status data of the person seeking help before seeking help; exemplarily, the medical and medication data may include whether the person seeking help sought medical treatment before seeking help, the medical diagnosis results, whether the medication was reused, the name of the medication, the continuous reuse time of the medication, and the efficacy of the medication.

[0100] In one embodiment, the patient status data may be obtained by any of the following methods:

[0101] The patient status data is obtained by asking relevant questions to the help seekers and integrating their responses to the relevant questions.

[0102] The medical treatment and medication data of the person seeking help are searched by associating the person seeking help's identity information, and the medical treatment and medication data and the help-seeking data are determined as patient status data; wherein the medical treatment and medication data may include the person seeking help's outpatient and hospitalization records, medical records, diagnosis certificates, and drug service status data.

[0103] It should be noted that the execution order of steps C1 and C2 can be adjusted successively or executed in parallel, which is not limited in the embodiment of the present application.

[0104] Step C3: Process the patient status data through the target network model to obtain the target rescue category.

[0105] In one embodiment, the target network model may be a multi-classification model. For example, after the target network model processes the patient status data, multiple rescue categories and the probabilities corresponding to each rescue category may be obtained. At this time, the rescue category with the highest probability may be determined as the target rescue category.

[0106] In one embodiment, the target rescue category may represent the category of the target object for which the request for help is made, wherein the target object may include the object for which the traumatic event and / or the accidental event in the aforementioned embodiments is directed. For example, when the traumatic event is a fracture event, the target object may include bones, and the target rescue category may be a bone injury category. When the accidental event is an indoor fire, the target object may include people, animals, and objects in the room, and the target request for help category may include an indoor fire extinguishing category and a fire escape category.

[0107] Step C4: determine the rescue guidance strategy corresponding to the target rescue category as the target rescue guidance strategy.

[0108] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, after determining the target network model and patient status data associated with the severity, the patient status data is processed through the target network model to obtain the target rescue category. In this way, since the factor of patient status data is added in the process of determining the target rescue category, the accuracy of the target rescue category can be improved; and, since the target rescue category is determined by the target network model associated with the severity, with the help of the data recognition, analysis and fusion capabilities of the target network model, the accuracy of the target rescue category and its correlation with the actual severity of the help seeker can be further improved; at the same time, determining the rescue guidance strategy corresponding to the target rescue category as the target rescue guidance strategy can improve the pertinence of the help guidance for the help seeker, thereby improving the efficiency of rescue guidance.

[0109] Based on the above embodiments, in the rescue guidance method provided in the embodiments of the present application, if the severity is greater than or equal to the target severity, obtaining the patient status data can be achieved by the following steps:

[0110] Step D1, obtaining the breathing state data and / or consciousness state data of the person seeking help.

[0111] In one embodiment, the target level may include the level of threat to the help-seeker's life and health greater than a specified threshold within a specified time period; illustratively, the target level may include a high-risk level; illustratively, the length of the specified time period may be determined based on historical data of medical assistance, the help-seeker's age, the help-seeker's environment, and the help-seeker's underlying disease level.

[0112] In one embodiment, the breathing status data may include whether the person seeking help is able to breathe normally; exemplarily, if the person seeking help is able to breathe normally, the breathing status data may include the person seeking help's breathing frequency and whether he is breathing deeply; exemplarily, if the person seeking help is not breathing, the breathing status data may be "no breathing".

[0113] In one embodiment, the consciousness state data may be descriptive data of the help-seeker's reaction state to external stimuli; exemplarily, the external stimuli may include stimuli in terms of light, temperature, air pressure, wind speed, and pressure; exemplarily, the above-mentioned descriptive data may include data such as the help-seeker's reaction time, reaction degree, and whether or not to react.

[0114] In one embodiment, the respiratory state data and the consciousness state data may be acquired through the statement of the person seeking help or the statement of the target object.

[0115] In one embodiment, it can be determined whether to collect respiratory status data and / or consciousness status data based on the help-seeking data; illustratively, if it is preliminarily determined based on the help-seeking data that the person seeking help is in a sick state, the respiratory status data and / or consciousness status data of the person seeking help can be obtained; for example, if the person seeking help is in an abnormal breathing state, the respiratory status data and consciousness status data can be obtained; for another example, if it is preliminarily determined based on the help-seeking data that the person seeking help is in a normal breathing state, the consciousness status data of the person seeking help can be obtained; for another example, if it is preliminarily determined based on the help-seeking data that the person seeking help is in a normal consciousness state, but the help-seeking data shows that the person seeking help has abnormal breathing, the respiratory status data can be obtained.

[0116] In one embodiment, when it is preliminarily determined that the person seeking help is in a sick state, if the breathing state data indicates that the person seeking help is at least in a state of continuous breathing, the consciousness state data can be obtained; illustratively, if the breathing state data indicates that the person seeking help is not in a breathing state, the consciousness state data may not be obtained; wherein continuous breathing may include continuous breathing represented by at least one of the respiratory rate, breathing depth, vital capacity and breathing mode in clinical medicine; wherein the breathing mode may include breathing through the nose or breathing through the mouth.

[0117] In one embodiment, if the breathing state data indicates that the person seeking help is in an intermittent breathing state, the consciousness state data may also be acquired.

[0118] Step D2: fusing the respiratory status data and / or consciousness status data and the help-seeking data to obtain patient status data.

[0119] In one embodiment, if it is preliminarily determined that the person seeking help is in a sick state, the respiratory state data and / or consciousness state data, and the help-seeking data may be fused to obtain the sick state data.

[0120] In one embodiment, the data in the respiratory state data and / or the data in the consciousness state data can be fused with the data related to the respiratory state and / or the consciousness state in the help-seeking data, respectively, to obtain the patient state data.

[0121] It should be noted that if the severity is less than the target severity, the patient status data may not be obtained through the above method.

[0122] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, if the severity is greater than the target degree, the respiratory status data and / or consciousness status data of the person seeking help are obtained. In this way, when the severity is greater than the target degree, the respiratory status data and / or consciousness status data of the person seeking help can be used to intuitively and accurately represent the actual state of the person seeking help; and, by fusing the respiratory status data and / or consciousness status data, as well as the help-seeking data, the patient status data can include the patient status of the person seeking help in multiple dimensions, thereby improving the comprehensiveness and accuracy of the patient status data.

[0123] Based on the above embodiments, in the rescue guidance method provided in the embodiment of the present application, obtaining the respiratory status data of the person seeking help can be achieved in the following ways:

[0124] Output the first prompt information to prompt the user to place the hand on the chest of the person seeking help, collect the ups and downs of the chest where the hand is placed; and determine the breathing state data based on the ups and downs.

[0125] In one embodiment, when the respiratory status data cannot be obtained, the first prompt information can be output in the form of voice or video to prompt the person seeking help or the target object to place their hands on the person seeking help's chest and collect the person seeking help's chest rise and fall status through a data acquisition device.

[0126] In one implementation, respiratory state data such as respiratory frequency and vital capacity may be determined based on the fluctuation frequency and fluctuation height in the fluctuation state.

[0127] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application outputs the first prompt information to prompt the user to place the hand on the chest of the person seeking help, thereby collecting the ups and downs of the chest where the hand is placed, and determining the respiratory state data based on the ups and downs. In this way, by collecting the ups and downs of the chest where the hand is placed to determine the respiratory state data, the accuracy of the respiratory state data can be improved; and by outputting the first prompt information to prompt the user to place the hand on the chest of the person seeking help, the probability of obtaining the respiratory state data can be increased.

[0128] Based on the above embodiments, in the rescue guidance method provided in the embodiments of the present application, obtaining the consciousness state data of the help-seeker can be achieved in the following ways:

[0129] Output a second prompt message to prompt the help-seeker to tap at least one body part, and collect the help-seeker's reaction state; determine the consciousness state data based on the reaction state.

[0130] In one embodiment, if the breathing status data indicates that the person seeking help is in a breathing state, the second prompt data may be output; if the breathing status data indicates that the person seeking help is in a non-breathing state, the second prompt data may not be output; exemplarily, the breathing status may include at least one of a normal breathing state, an abnormal continuous breathing state, and an intermittent apnea state.

[0131] In one embodiment, at least one body part may be determined based on the help-seeking data and the breathing status data; illustratively, the degree of harm caused by tapping at least one body part to the help-seeker may be less than a specified degree.

[0132] In one embodiment, the second prompt information may include prompt information of patting the help-seeker's shoulder or pressing the help-seeker's arm in the form of voice or video, so as to prompt the target object or the help-seeker to apply external stimulation to the help-seeker, and obtain or collect the help-seeker's stimulation response, thereby obtaining the reaction state.

[0133] In one embodiment, the reaction states within a certain period of time may be statistically integrated to determine the consciousness state data.

[0134] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application outputs a second prompt information to prompt the person seeking help to pat at least one body part, thereby increasing the probability of the person seeking help to respond in the current state, and then improving the accuracy of the reaction state. On this basis, determining the consciousness state data based on the reaction state can also improve the accuracy of the consciousness state data, and can also increase the probability of obtaining the consciousness state data.

[0135] Based on the foregoing embodiments, in the rescue guidance method provided by the embodiments of the present application, if the severity is greater than or equal to the target severity, the target network model includes a first model; the first model includes a first mapping module, a first association module, a first extraction module and a first classification module.

[0136] Accordingly, the patient status data is processed through the target network model to obtain the target rescue category, which can be achieved through the following steps:

[0137] Step E1: Mapping the patient status data through a first mapping module to obtain a first mapping result.

[0138] In one implementation, the patient status data may be segmented to obtain a segmentation result, which is then input into the first mapping module.

[0139] In one implementation, the word segmentation results may be numerically processed to obtain a data stream corresponding to the patient status data, and then the data stream is input into the first mapping module.

[0140] In one embodiment, the first mapping module can map the word segmentation result or data stream to a higher dimensional vector space, thereby providing data dimension support for subsequent context association and feature extraction. That is, the first mapping result can be a multidimensional matrix, and the dimension of the first mapping result can be higher than the dimension of the word segmentation result or data stream.

[0141] Figure 5 This is a schematic diagram of the structure of the first model provided in the embodiment of the present application. Figure 5 As shown, d1, d2, ..., d N The first mapping module 501 may include an Embedding, which is used to map d1, d2, ..., d N Perform vector mapping respectively to obtain vector sets e1, e2, ..., e N ; Exemplarily, the first mapping result may include e1, e2, ..., e N Wherein, N can be an integer greater than 3.

[0142] Step E2: Process the first mapping result through a first association module to obtain a first association result.

[0143] In one implementation, the first association module is used to extract the context association relationship in the first mapping result, that is, the first association result may include the context association relationship between each vector data in the first mapping result.

[0144] In one embodiment, the first association module may reduce the dimension of the first mapping result, that is, the dimension of the first association result may be smaller than the dimension of the first mapping result, which may represent the deep features contained in the patient status data.

[0145] For example, Figure 5 As shown, the first association module 502 may include an Attention module, which extracts the association relationship between the N vectors included in the first mapping result to obtain K association vectors h1, h2, ..., h K , where K can be a positive integer less than N.

[0146] Step E3: Process the first association result through the first extraction module to obtain a first extraction result.

[0147] In one embodiment, the first extraction module is used to extract standardized medical patient features contained in the first association result.

[0148] like Figure 5As shown, the first extraction module 503 may include N extraction units, and the extraction units may be in a cascade connection relationship, and the output of each extraction unit may be a vector.

[0149] Step E4: classify the first extraction result by using a first classification model to obtain a first guidance category.

[0150] In one embodiment, the first classification model can calculate the first extraction result including N extraction results, thereby obtaining N probabilities corresponding to the N extraction results. Then The guidance category corresponding to the maximum value in is determined as the first guidance category.

[0151] For example, Figure 5 As shown, the first classification model 504 may include softmax, which is capable of performing multi-classification on N extraction results.

[0152] Step E5: Determine the first guidance category as the target guidance category.

[0153] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, if the severity is greater than or equal to the target degree, the target network model is the first model. In this way, targeted and differentiated processing of patient status data with a severity greater than the target degree is achieved through the first model; and, by mapping the patient status data through the first mapping module, the spatial characteristics of the first mapping result can be enhanced, so that the first association result obtained by processing the first mapping result by the first association module can include the deep-level association characteristics between different data in the first mapping result; on this basis, by processing the first association result through the first extraction module, the richness and accuracy of the medical features in the first extraction result can be improved, thereby improving the accuracy of the first guidance category obtained by classifying the first extraction result by the first classification module.

[0154] Based on the aforementioned embodiments, in the rescue guidance method provided in the embodiments of the present application, the first extraction module includes a feed forward network (Feed Forward, FF) module comprising N extraction units, and the N extraction units are respectively connected to the first classification module; N is an integer greater than 3.

[0155] In one embodiment, a forward feedback data connection may be established between the input and output of the N extraction units, such as Figure 5 As shown, the input data of the second extraction unit includes the extraction result of the first extraction unit and the classification result associated with the first extraction unit output by the first classification module.

[0156] Accordingly, performing feature extraction processing on the first association result by the first feature extraction module to obtain the first extraction result can be achieved by the following steps:

[0157] Step F1: classify the extraction result output by the n-1th extraction unit in the first extraction module using a first classification model to obtain a first classification probability.

[0158] Here, n is an integer greater than 1 and less than N.

[0159] In one embodiment, the extraction results output by the first extraction unit to the extraction results output by the n-1th unit can be classified by the softmax included in the first classification model to obtain a first probability set containing n-1 probabilities, and the probability corresponding to the extraction result of the n-1th extraction unit in the first probability set is determined as the first classification probability.

[0160] Step F2: Perform feature extraction on the first classification probability and the extraction result of the n-1th extraction unit through the nth extraction unit to obtain the extraction result of the nth extraction unit.

[0161] Exemplarily, the input data of the nth extraction unit may include the extraction result output by the (n-1)th extraction unit and the first classification probability.

[0162] Accordingly, the first extraction result is classified by the first classification module to obtain the first guidance category, which can be achieved in the following manner:

[0163] The extraction results of the first extraction unit to the extraction results of the Nth extraction unit are processed by the first classification module to obtain N classification probabilities; based on the N classification probabilities, a first guidance category is determined.

[0164] Exemplarily, the softmax included in the first classification module can be used to perform multi-classification processing on the extraction results of the first extraction unit to the extraction results of the Nth extraction unit, so as to obtain the probabilities corresponding to the above-mentioned N extraction results, that is, N classification probabilities, and determine the guidance category corresponding to the highest probability among the N classification probabilities as the first guidance category.

[0165] Exemplarily, softmax can process the extraction results of the first extraction unit to the extraction results of the Nth extraction unit through formula (1) to obtain the classification probability of the tth output of softmax:

[0166]

[0167] Among them, O tis a set of extraction results from the first extraction unit to the Nth extraction unit, and t is an integer greater than 1; Q t The value of can be -∞ or 0, where if Q t If it has been output by softmax before t-1 times, that is, it has been predicted, then its value can be -∞. t If it has not been output by softmax in the previous t-1 times, that is, it has been predicted, its value can be 0.

[0168] Through the above Q t The limitation of the value of can weaken the influence of the classification probability obtained by historical calculation on the classification probability of the t-th output, thereby reducing the correlation and coupling between the classification probabilities calculated at different times.

[0169] Correspondingly, with the classification probability The first guidance category It of the t-th output of the corresponding first model can be calculated by formula (2):

[0170]

[0171] From the above, it can be seen that in the rescue guidance method provided by the embodiment of the present application, the first extraction module includes a forward network module including N extraction units, and the N extraction units are respectively connected to the first classification module, so that the first classification module can classify the extraction results of each extraction unit in real time; and after the first classification model classifies the extraction results output by the n-1th extraction unit to obtain the first classification probability, the nth extraction unit performs feature extraction on the first classification probability of the nth extraction unit and the extraction result of the n-1th extraction unit, so that the extraction result of the nth extraction unit obtained through the above steps can carry the extraction result of the n-1th extraction unit and the relevant information of the first classification probability, thereby enhancing the correlation between the extraction result of the nth extraction unit and the extraction result of the n-1th extraction unit and the first classification probability; at the same time, after obtaining N classification probabilities through the first classification model, the first guidance category is determined based on the N classification probabilities, thereby realizing multi-classification processing of the N extraction results of the N extraction units, thereby improving the accuracy of the first guidance category obtained by the above method.

[0172] Figure 6 A flowchart of a target rescue guidance strategy for determining a high-risk level provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the process may include the following steps:

[0173] Step 601: Determine whether there is breathing.

[0174] For example, it can be determined whether the person seeking help is breathing by asking the person seeking help, or by collecting breathing data in real time through a portable device worn by the person seeking help.

[0175] Exemplarily, if the person seeking help is in a breathing state, respiratory status data such as respiratory rate can be obtained, and step 602 can be executed; if the person seeking help is in a non-breathing state, the respiratory status data can be marked as "non-breathing", and step 603 can be executed; if it is impossible to determine whether the person seeking help is still breathing, step 604 can be executed.

[0176] Step 602: Determine whether the person is conscious.

[0177] For example, whether the help seeker is conscious can be determined by asking the help seeker questions and the help seeker's response status.

[0178] Exemplarily, if the person seeking help is in a conscious state, the consciousness state data can be obtained and step 603 is executed; if the person seeking help is in an unconscious state, the consciousness state data is marked as "unconscious" and step 603 is executed; if it is impossible to determine whether the person seeking help is in a conscious state, step 605 is executed.

[0179] Step 603: Determine the target rescue guidance strategy through the first model.

[0180] Exemplarily, the method provided in the aforementioned embodiment can be used to first integrate the respiratory status data, consciousness status data and help-seeking data to obtain patient status data, and then process the patient status data through the various modules in the first model to determine the first guidance category, and determine the rescue guidance strategy corresponding to the first guidance category as the target rescue guidance strategy.

[0181] Step 604: Assist in determining whether there is breathing.

[0182] Exemplarily, the method of outputting the first prompt information provided in the aforementioned embodiment can be used to assist in determining whether the person seeking help is breathing by observing the rise and fall of the chest.

[0183] Exemplarily, step 604 and step 601 may be repeatedly switched for a specified number of times to increase the probability of obtaining respiratory state data.

[0184] Step 605: Assist in determining whether the person is conscious.

[0185] Exemplarily, the method for outputting the second prompt information provided in the aforementioned embodiment can be used to obtain the reaction state of the person seeking help when at least one body part is tapped, and to assist in determining whether the person seeking help is conscious based on the reaction state.

[0186] Exemplarily, step 602 and step 605 may be repeatedly switched for a specified number of times to increase the probability of obtaining consciousness state data.

[0187] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application, when the severity of the help-seeker is at a high-risk level, can improve the success rate of obtaining respiratory status data and consciousness status data by assisting in judging whether the person is breathing and conscious, thereby laying a data foundation for determining the target rescue guidance strategy through the first model, thereby enabling the targeted rescue guidance strategy to be targeted and improving the efficiency and success rate of rescue guidance.

[0188] Based on the aforementioned embodiments, in the rescue guidance method provided in the embodiments of the present application, the first model is obtained by training the first model of the initial state based on sample data, and the sample data includes a high-risk rescue guidance identifier.

[0189] In one embodiment, the sample data may include patient status data of various severity levels, and may also include high-risk rescue guidance strategies corresponding to the patient status data.

[0190] Table 2

[0191] Serial number Text Regular Expression 1 ? <! [shallow|mild])(severe|serious)? coma|(unconscious(?!disorder))|loss of consciousness(?!history) 2 Cardiac arrest | cardiac arrest | cardiac arrest | cardiac arrest | no heartbeat | myocardial infarction 3 Respiratory arrest|Respiratory arrest|Respiratory pause|No breathing (?! Distress|Difficulty) 4 Airway obstruction | suffocation | failure 5 Severe respiratory distress | Acute respiratory distress | Acute dyspnea | Severe impaired consciousness 6 ((Complex injury|Multiple injuries).*(?<![No|No|Not])Severe breathing difficulties)

[0192] In one embodiment, a variety of respiratory status data, consciousness status data, and other vital signs data obtained during the first aid process can be processed using a regular matching method to obtain sample data. For example, a text regular expression for high-risk patient guidance types can be set based on prior medical knowledge, and the corpus data obtained during the first aid process can be labeled with high-risk guidance types based on the text regular expression to obtain sample data.

[0193] Table 2 is an example statistical result of the regular expression provided in the embodiment of the present application.

[0194] For example, if the first aid help corpus obtained after conversion is "peanuts stuck in organs, suffocating", then the text regular expression shown in Table 2 is used to perform regular matching on the above help corpus, and it can be determined that the help corpus is a high-risk rescue guidance type with sequence number 4. If the help corpus does not match any expression in the text regular expression, it can be determined that the help is not expected to be a high-risk rescue guidance type.

[0195] At the same time, due to the richness of natural language, the above text regular expression may not cover all help-seeking corpora, or the text regular expression may not match the help-seeking corpora correctly. In order to reduce the probability of the above problems, medical professionals can manually correct the matching results of the help-seeking corpora and use the manually corrected data as sample data.

[0196] In one implementation, the high-risk rescue guidance identifier may include identifiers of high-risk rescue guidance categories corresponding to various help-seeking corpora; illustratively, the high-risk rescue guidance identifier may be embodied in the form of digital codes, characters, and text sets.

[0197] Accordingly, the rescue guidance method provided in the embodiment of the present application may further include the following steps:

[0198] Step G1: Map the sample data through a first mapping module to obtain first vector data.

[0199] Exemplarily, the mapping process of the sample data by the first mapping module may be the same as the mapping process of the patient status data by the first mapping module in the aforementioned embodiment.

[0200] Exemplarily, the first vector data may have the same meaning as the first mapping result in the aforementioned embodiment.

[0201] Figure 7 A schematic diagram of the structure of the first model in the initial state of training provided in the embodiment of the present application, such as Figure 7 As shown, the sample data can be Figure 7 x1, x2, ..., x shown in N , z1, z2, ..., z N It can be the first vector data.

[0202] Step G2: Perform context association on the first vector data through a first association module to obtain first context data.

[0203] Exemplarily, the context association performed by the first association module on the first vector data may be the same as the association processing performed by the first association module on the first mapping result data in the aforementioned embodiment.

[0204] Exemplarily, the first context data may have the same meaning as the first association result in the aforementioned embodiment.

[0205] Exemplarily, the first context data may be Figure 7 C1, C2, ..., C K .

[0206] Step G3: Process the high-risk rescue guidance identifier and the first context data through the first extraction module and the first classification module to obtain a prediction identifier.

[0207] Exemplarily, the input data of the first extraction module may include the first context data and a high-risk rescue guidance identifier.

[0208] Exemplarily, the above processing process may be the same as the process of obtaining the first guidance category in the aforementioned embodiment.

[0209] Exemplarily, the prediction identifier may have the same meaning as the first guidance category in the aforementioned embodiment, and is used to indicate the prediction result of the high-risk rescue guidance category identifier corresponding to the sample data.

[0210] Step G4: Based on the matching degree between the prediction mark and the high-risk rescue guidance mark, at least the weight parameter of the first extraction module is adjusted to obtain the first model.

[0211] Exemplarily, the weight parameters of each module in the first model in the initial state may be adjusted according to the degree of matching between the prediction identifier and the high-risk rescue guidance identifier, thereby obtaining the first model.

[0212] For example, the above matching degree can be reflected by the cross loss entropy between the prediction mark and the high-risk rescue guidance mark, which can be calculated by formula (3):

[0213]

[0214] Among them, Loss is the cross loss entropy between the prediction mark and the high-risk rescue guidance mark, N is the number of prediction marks and high-risk rescue guidance marks, y i is the i-th high-risk rescue guidance mark, The i-th prediction identifier.

[0215] Exemplarily, when the value of Loss is less than or equal to the target loss, training can be stopped to obtain the first model.

[0216] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, the sample data includes a high-risk rescue guidance identifier. Thus, based on the sample data and the high-risk rescue guidance identifier, at least the weight parameters of the first extraction model in the first model in the initial state are adjusted to obtain the first model, so that the first model can accurately classify the rescue guidance type of the patient status data it obtains.

[0217] Based on the aforementioned embodiments, in the rescue guidance method provided in the embodiments of the present application, the first extraction module includes a first FF network module comprising N extraction units, and the N extraction units are respectively connected to the first classification module; N is an integer greater than or equal to 3.

[0218] Accordingly, the high-risk rescue guidance mark and the first context data are processed by the first extraction module and the first classification module to obtain the prediction mark, which can be achieved by the following steps:

[0219] Step H1: weight the n-1th high-risk rescue guidance indicator with a first weight to obtain a first result.

[0220] Among them, the n-1th high-risk rescue guidance identifier includes the high-risk rescue guidance identifier of the n-1th data in the sample data; n is an integer greater than 1 and less than N.

[0221] In one embodiment, the first weight may be predetermined; for example, Figure 7 As shown, the first weight may be 1-P, where P is a number greater than 0 and less than 1.

[0222] For example, the first result may be 1-P and y n The product of .

[0223] Step H2: weight the n-1th guidance category probability output by the first classification module with a second weight to obtain a second result.

[0224] Among them, the n-1th guidance category probability is at least associated with the output data of the n-1th extraction unit in the first preceding network, and the output data of the n-1th extraction unit is associated with the nth data in the sample data.

[0225] In one embodiment, the sum of the first weight and the second weight associated with the same classification unit may be 1, that is, when the first weight is P, the second weight may be 1-P.

[0226] For example, the n-1th guidance category probability can be Figure 7 As shown in The second result can be 1-P and The product of .

[0227] Step H3: Process the output data of the nth extraction unit, the n-1th extraction unit, the first result and the second result through the nth extraction unit in the FF network module to obtain the output data of the nth extraction unit.

[0228] Exemplarily, the nth extraction may perform feature extraction on the output data of the (n-1)th extraction unit, the first result, and the second result, so as to obtain the output data of the nth extraction unit.

[0229] Step H4: classify the output data from the first extraction unit to the output data of the Nth extraction unit through the first classification module to obtain a prediction mark.

[0230] Exemplarily, the first classification module performs multi-classification processing on the output data of the first extraction unit to the output data of the Nth extraction unit to obtain N probabilities, and determines the identifier corresponding to the maximum value among the N probabilities as the predicted identifier.

[0231] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, the nth high-risk rescue guidance identification is weighted with a first weight to obtain a first result, and the n-1th guidance category probability output by the first classification module is weighted with a second weight to obtain a second result, and the first result and the second result are input into the nth extraction unit to obtain the output data of the nth extraction unit. In this way, by inputting the second result into the nth extraction unit, the feature extraction result of the n-1th extraction unit and the feature extraction process of the nth extraction unit are correlated with each other; and by inputting the first result into the nth extraction unit, targeted adjustment of the feature extraction parameters of each extraction unit based on the high-risk rescue guidance identification in the sample data can be achieved; at the same time, by adjusting the first weight and the second weight, the training process of the first network in the initial state can be effectively controlled, thereby enhancing the controllability of the above-mentioned training process.

[0232] Based on the foregoing embodiment, in the rescue guidance method provided in the embodiment of the present application, the sample data is mapped by the first mapping module to obtain the first vector data, which can be implemented by the following steps:

[0233] Step J1, count the high-risk rescue guidance marks, and obtain the mth mark number corresponding to the high-risk rescue guidance marks of the mth guidance category.

[0234] Wherein, m is an integer greater than or equal to 1 and less than or equal to M; M is the number of rescue guidance categories included in the high-risk rescue guidance identification.

[0235] In one implementation, the high-risk rescue guidance identifiers respectively included in all the data in the sample data may be classified to obtain the first high-risk category identifier to the Mth high-risk category identifier.

[0236] In one embodiment, the high-risk rescue guidance identifier in the sample data may include a guidance category identifier. In this way, through the guidance category identifier, rapid classification of each high-risk rescue guidance identifier can be achieved, and the mth identifier number can be obtained based on the classification result.

[0237] Step J2: Arrange the first identification quantity to the Mth identification quantity in descending order to obtain a sorting result.

[0238] Step J3: Arrange the data associated with the first guidance category to the data associated with the Mth guidance category in the sample data based on the sorting result to obtain an arrangement result.

[0239] Exemplarily, the data respectively associated with different guidance categories in the sample data may be rearranged based on the sorting result to obtain an arrangement result.

[0240] Step J4: Map the arrangement result through the first mapping module to obtain first vector data.

[0241] In actual applications, the probability of rescue guidance categories with strong correlation appearing at the same time is higher than the probability of rescue guidance categories with weak correlation appearing at the same time. For example, the probability of drowning guidance and cardiopulmonary resuscitation guidance appearing at the same time is higher than the probability of drowning guidance and childbirth guidance appearing at the same time. Therefore, the probability of a drowning guidance case also having a cardiopulmonary resuscitation guidance category is higher.

[0242] Based on the above situation and actual rescue guidance needs, the embodiment of the present application arranges the high-risk rescue guidance categories appearing in the sample data in descending order, and arranges the data of different guidance categories in the sample data according to the sorting result, so as to extract the strong correlation between a large number of high-risk rescue guidance categories; and, maps the arrangement result through the first mapping module to obtain the first vector data, so that the correlation between the data in the above arrangement result can guide the training process of the first model in the initial state, improve the convergence speed of the above training process, shorten the training time, and also can improve the classification accuracy of the first model.

[0243] Based on the above embodiments, in the rescue guidance method provided in the embodiments of the present application, if the severity is less than the target severity, the patient status data is obtained, which can be achieved by the following steps:

[0244] Step K1, obtain a target question list including patient status questions.

[0245] In one embodiment, a severity level less than a target level may indicate that the threat to the life and health of the person seeking help from the current state is less than a preset level; illustratively, the person seeking help may be at a normal level at this time.

[0246] In one embodiment, patient status questions may include various types of patient problems that a caller may encounter.

[0247] In one implementation, the patient status questions may include questions corresponding to the current patient status of the person seeking help.

[0248] In one embodiment, the patient status questions may include questions about at least one symptom manifestation, the subjective feeling of the person seeking help, the duration, and medication history.

[0249] In one embodiment, the target question list may include multiple patient status questions, and these patient status questions may cover multiple levels or dimensions of patient status.

[0250] In one implementation, the target question list may be obtained in the following manner:

[0251] When the severity is less than the target severity, a question set corresponding to the patient status of the help seeker is determined according to the severity and the help seeker's help data, and the question set is determined as the target question list.

[0252] Step K2: Obtain the target answer data of the help seeker for the questions in the target question list.

[0253] In one implementation, the target reply data may include voice data, text data, image data, and video data sent by the help seeker or target object.

[0254] In one implementation, the target response data may be obtained by:

[0255] The answer data directly input by the help seeker or target object is obtained, and the above answer data is integrated into the target answer data.

[0256] A selection operation of at least one option provided by the help seeker or target object for the target question list is obtained, a selection result is obtained, and the selection result is determined as target answer data.

[0257] Step K3: Determine the patient status data based on the target question list and the target answer data.

[0258] In one embodiment, the patient status data may be determined by:

[0259] Extract the stem content from the questions in the target question list, calibrate the answer data corresponding to the above questions in the target answer data to obtain valid answer data, and integrate the stem content with the valid answer data to determine the patient status data.

[0260] From the above, it can be seen that in the rescue guidance method provided by the embodiment of the present application, if the severity is less than the target severity, a target question list including patient status questions is first obtained. In this way, the target question list can comprehensively and flexibly reflect the types and dimensions of information that need to be collected for the help seeker; on this basis, the target answer data of the help seeker for the questions in the target question list is obtained. Through the above target answer data, the current real status of the help seeker can be obtained more comprehensively and accurately; and based on the target question list and the target answer data, the patient status data is determined, so that the patient status data can comprehensively display the status of the help seeker from multiple dimensions and multiple levels, thereby improving the accuracy of the patient status data.

[0261] Based on the foregoing embodiments, in the rescue guidance method provided by the embodiments of the present application, the target question list includes a first question list associated with the initial plan.

[0262] In one embodiment, the initial plan includes a rescue category obtained by preliminarily classifying the help seekers.

[0263] In one embodiment, the first list of questions may include a list of multiple questions associated with the initial plan.

[0264] Accordingly, obtaining the target question list of patient status questions can be achieved through the following steps:

[0265] Step L1: classify the help-seeking data through the emergency plan classification model to obtain an initial emergency plan.

[0266] In one embodiment, the emergency plan classification model may be a binary classification model. In this case, the initial emergency plan may indicate whether the person seeking help needs immediate emergency assistance.

[0267] In one embodiment, the emergency plan classification model may be a multi-classification model. In this case, the initial emergency plan may characterize the smaller-granularity rescue category to which the help seeker actually belongs.

[0268] Exemplarily, the structure of the plan classification model may be the same as the structure of the critical level classification model provided in the aforementioned embodiment, and will not be described in detail here.

[0269] According to the actual needs of plan classification, plan sample data can be set for the plan classification model, and the above data can include multiple plan identifiers and plan description data; illustratively, the above plan identifiers and plan description data can be embodied in the form of key-value pairs, such as the key-value pair can be Z1: Z2, wherein Z1 can be a plan identifier expressed in digital form, and Z2 can be plan description data expressed in text form, and the above plan sample data can be as follows:

[0270] 0: heat / cold exposure, 1: stroke, 2: heart problems, 3: electrical injuries / lightning injuries, 4: animal bites / attacks, 5: falls from heights, 6: pregnancy / delivery / miscarriage, 7: coma / syncope, 8: transportation accidents, 9: drowning / diving accidents, 10: airway obstruction, 11: burns / explosions, 12: chest pain, 13: eye problems / injuries, 14: carbon monoxide / inhalants / hazardous materials, 15: convulsions / seizures, 16: cardiac or respiratory arrest or death, 17: trauma / bleeding / wounds, 18: abdominal pain, 19: difficulty breathing.

[0271] The above data are all in the form of Z1:Z2, and the initial plan classification model can be trained using the above data to obtain a plan classification model; illustratively, the plan classification model can output the probability that the help request data belongs to at least one plan, and determine the plan with the highest probability as the initial plan.

[0272] Step L2: Obtain a first list of questions associated with the initial plan.

[0273] In one embodiment, the initial plan and the first question list may have an association relationship, so that after the initial plan is determined, the first question list associated with the initial plan category may be determined based on the initial plan category and the above association relationship.

[0274] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, the help-seeking data is classified through the plan classification model to obtain the initial plan. In this way, the preliminary classification of the help-seeking data can be achieved, thereby providing plan category conditions for obtaining the first question list associated with the initial plan, and also improving the accuracy of the first question list.

[0275] Based on the foregoing embodiment, in the rescue guidance method provided in the embodiment of the present application, the target question list includes a second question list.

[0276] In one implementation, at least some of the questions in the second question list may be different from at least some of the questions in the first question list.

[0277] In one implementation, the plan corresponding to the second list of questions may be different from the initial plan corresponding to the first list of questions.

[0278] Accordingly, obtaining a target question list including patient status questions can be achieved by the following steps:

[0279] Step M1: classify the help-seeking data through the emergency plan classification model to obtain an initial emergency plan.

[0280] Step M2: Obtain response data for the first list of questions associated with the initial plan.

[0281] Step M3: If the reply data is associated with at least one question outside the first question list, determine that the plan associated with at least one question is a jump plan, and determine that the question list associated with the jump plan category is a second question list.

[0282] Exemplarily, if the answer data is not associated with a question outside the first question list, it can be determined that the target question list is the first question list in the aforementioned embodiment.

[0283] In one embodiment, the number of plans included in the jump plan class may be at least one. For example, if the above-mentioned reply data is associated with the question list of the first plan, and the reply data for the question list of the first plan is associated with the question list of the second plan, then the jump plan may include the first plan and the second plan.

[0284] In one embodiment, a list of questions associated with multiple plans can be pre-established, and the above-mentioned question list can include a plan category identifier, a plan name, a question identifier, a question name, an answer identifier, an answer content, whether it is a single choice, and an associated question identifier, etc. Taking the chest pain plan category as an example, the above-mentioned question list can be shown in Table 3. In Table 3, whether it is a single choice indicates whether the current question is a single choice question, and the associated question identifier indicates the identifier of the next question associated with the answer to the current question, wherein, if the associated question identifier matches the question identifier included in the current plan, the plan jump is not executed, and if the associated question identifier does not match the question identifier included in the current plan, the plan jump is executed. For example, when the associated question identifier in Table 3 is 7 or 9, the question identifiers included in the chest pain plan do not include question identifiers of 7 or 9, and the plan jump can be executed.

[0285] Table 3

[0286]

[0287] For example, Figure 8 A schematic diagram of the process flow of the plan jump provided in the embodiment of the present application, such as Figure 8 As shown, after obtaining the help-seeking data of slight crooked mouth corner 801, the help-seeking data can be multi-classified through the plan classification model, so as to determine that the initial plan is the cerebral infarction plan 802, and obtain the first question list of the cerebral infarction plan 802, such as Figure 8 As shown in the figure, questions such as whether the person is awake, breathing, and having chest pain are asked. When the person seeking help answers "yes" to the question of whether he has chest pain, the answer is associated with the chest pain plan 803. At this time, you can jump from the initial plan classification to the chest pain plan 803, and obtain the list of questions included in the chest pain plan 803. Then, based on the response data of the above question list, determine whether to execute the plan jump.

[0288] Fig. 9 Another schematic diagram of the process of plan jump provided in the embodiment of the present application is as follows: Fig. 9 As shown, in the case where the help-seeking data 901 is fainting at home, the help-seeking data 901 is subjected to multi-classification processing through the plan classification model, and the initial plan category 902 can be determined to be coma / fainting. However, according to the response data of the first question list of the coma / fainting plan, it can be determined that the associated question list 903 includes problems such as movement disorders before fainting, gas poisoning, and increased heart rate before fainting. Therefore, a jump plan 904 including convulsions / seizures, carbon monoxide / inhalants / dangerous items and heart can be obtained.

[0289] pass Figures 8 to 9The plan jump process shown provides an initial, rough plan classification for the help seeker to a refined plan classification process, thereby improving the comprehensiveness and accuracy of the plan classification.

[0290] From the above, it can be seen that in the rescue guidance method provided in the embodiment of the present application, after the help-seeking data is classified by the plan classification model to obtain the initial plan and the reply data for the first question list associated with the initial plan is obtained, if the reply data is associated with at least one question outside the first question list, the plan associated with at least one question is determined to be a jump plan, and the question list associated with the jump plan is determined to be the second question list. As a result, not only the diversified, refined and standardized plan classification processing of the help-seekers is achieved, but also the accuracy and comprehensiveness of the target question list can be improved.

[0291] Based on the foregoing embodiments, in the rescue guidance method provided by the embodiments of the present application, when the severity is less than the target severity, the target network model includes a second model; the second model includes a second mapping module, a second association module, a second extraction module and a second classification module.

[0292] Accordingly, the patient status data is processed through the target network model to obtain the target rescue category, which can be achieved through the following steps:

[0293] Step T1, obtain the target plan associated with the target problem list.

[0294] In one implementation, the target plan may include an initial plan or a jump plan.

[0295] Step T2: Process the plan name, help-seeking data, target question list and target answer data of the target plan through the second mapping module to obtain a second mapping result.

[0296] Fig.10 This is a schematic diagram of the structure of the second model provided in the embodiment of the present application. Fig.10 As shown, the second model may further include an input module 1001, which is used to input the help-seeking data, the first plan-related data, the second plan-related data to the M-1 plan-related data into the second mapping module 1002; wherein the first plan-related data may include a list of questions and reply data associated with the first plan in the jump plan, and the second plan-related data may include a list of questions and reply data associated with the second plan in the jump plan; the M-1 plan-related data may include a list of questions and reply data associated with the M-1 plan in the jump plan; wherein M may be an integer greater than 2.

[0297] Exemplarily, the question list contained in the above-mentioned first plan-associated data to the M-1th plan-associated data may be the target question list in the aforementioned embodiment, and the answer data contained in the above-mentioned first plan-associated data to the M-1th plan-associated data may be the target answer data in the aforementioned embodiment.

[0298] In one embodiment, the target question list and the target answer data are segmented according to the plan category, and the following can be obtained: Fig.10 The first plan-related data and the second plan-related data are shown.

[0299] In one implementation, the plan name may include an identifier of the plan, such as chest pain in the aforementioned embodiment.

[0300] In one embodiment, the help-seeking data and the data associated with different plans can be respectively input into different mapping neurons of the second mapping module, so as to obtain the input data v1, v2, ..., v M .

[0301] In one embodiment, the second mapping module maps v1, v2, ..., v M The second mapping result obtained after processing can have a dimension higher than v1, v2..., v M The vector set q1, q2, ..., q M .

[0302] In one implementation, the second mapping module may include embedding.

[0303] Step T3: context-associate the second mapping result through a second association module to obtain a second association result.

[0304] In one embodiment, the second association module 1003 may include Attention, which is used to extract context data in the second mapping result, so as to obtain deep features contained in the second mapping result. The second mapping result may be as follows: Fig.10 p1, p2, ..., p shown in K .

[0305] Step T4: Process the second association result through the second extraction module and the second classification module to obtain a second guidance category.

[0306] In one embodiment, the second extraction module 1004 may include an FF network module, which may include N feature extraction units, and the input data of the nth feature extraction unit may include probability data obtained after the second classification module processes the n-1th feature extraction unit, and a second association result.

[0307] In one implementation, the second classification module 1005 may classify the output data of the N feature extraction units to obtain a plurality of classification probabilities w1, w2, ..., w N , and w1, w2, ..., w N The guidance category corresponding to the maximum value in is determined as the second guidance category.

[0308] Step T5: Determine the second guidance category as the target rescue category.

[0309] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application, after obtaining the target plan associated with the target question list, processes the plan name, help data, target question list and target answer data of the target plan through the second mapping module to obtain a second mapping result, so that the second mapping result can include the comprehensive patient characteristics contained in the target question list, target answer data and help data; on this basis, the second association result obtained by processing the second mapping result through the second association module can comprehensively include the deep features in the second association result; and, the second association module is processed by the second extraction module with the second classification module to obtain and determine the second guidance category as the target rescue category, thereby improving the pertinence and accuracy of the target rescue category.

[0310] Fig.11 A schematic diagram of a process for determining a common patient guidance strategy provided in an embodiment of the present application. Fig.11 As shown, the process may include the following steps:

[0311] Step 1101: Obtain chief complaint data.

[0312] Illustratively, the main complaint data may be the main complaint content in the aforementioned embodiment.

[0313] Step 1102: preprocessing.

[0314] Exemplarily, the main complaint data can be processed by the method provided in the aforementioned embodiment to obtain the help-seeking data.

[0315] Step 1103: Plan processing.

[0316] Exemplarily, if it is determined through the method provided in the aforementioned embodiment that the critical condition of the person seeking help is less than the target condition, the emergency plan classification model can be used to perform emergency plan classification processing on the help-seeking data to obtain an initial emergency plan.

[0317] Step 1104: Obtain a target question list.

[0318] Exemplarily, the first question list corresponding to the initial plan can be obtained first, and the response data for the first question list can be obtained. If the response data is associated with at least one question outside the first question list, it can be determined that the plan associated with at least one question is a jump plan.

[0319] Step 1105: Determine whether to jump to the plan.

[0320] Exemplarily, if the jump plan has been made, step 1104 may be continued to be executed to obtain the second question list associated with the jump plan and the answer data corresponding to the second question list.

[0321] Exemplarily, if the jump plan ends or the jump plan is not executed, step 1106 can be executed.

[0322] Step 1106: Generate a second rescue guidance strategy.

[0323] Exemplarily, the first question list and its answer data corresponding to the initial plan, the second question list and its answer data corresponding to the jump plan, and the help-seeking data can be input into the second model. After being processed by various modules in the second model, the second guidance category is obtained, and the rescue guidance strategy corresponding to the second guidance category is determined as the target rescue guidance strategy.

[0324] From the above, it can be seen that in the embodiment of the present application, when the person seeking help is in a normal rescue state, through the plan jump operation and obtaining the initial plan and the target problem list associated with the jump plan, the patient status data of the person seeking help can be deeply mined, thereby improving the pertinence and accuracy of the second rescue guidance strategy.

[0325] In the case of limited medical resources, if multiple help-seekers apply for assistance guidance at the same time, in order to give priority to allocating limited medical resources to help-seekers in high-risk conditions, the embodiment of the present application can also set a level identification for each help-seeker.

[0326] Exemplarily, the level identification can be represented by different colors, which are specifically used to distinguish the severity of the situations of different help-seekers; exemplary, the severity levels can include high-risk level, critical level and non-emergency level. In actual applications, for help-seekers at high-risk levels, not only can they be provided with the highest priority rescue guidance, but also medical resources such as ambulances can be dispatched to them on a limited basis.

[0327] Fig.12 This is a schematic diagram of the process of setting the level identification provided in the embodiment of the present application. Fig.12 As shown, the process may include the following steps:

[0328] Step 1201: Obtain help request data.

[0329] Step 1202: Classify high-risk types.

[0330] Exemplarily, the help-seeking data may be subjected to binary classification processing using the critical level classification model provided in the aforementioned embodiment, thereby determining whether the severity of the help-seeker is greater than the target severity, that is, determining whether the help-seeker is of the high-risk type.

[0331] Step 1203: Determine whether it is a high-risk type.

[0332] Exemplarily, if the person seeking help is of a high-risk type, step 1204 is executed; if the person seeking help is not of a high-risk type, step 1205 is executed.

[0333] For example, if the severity of the illness of the help-seeker is different, the urgency of the rescue guidance for the help-seeker may also be different. Therefore, in order to more intuitively reflect the rescue level of different help-seekers, different color labels can be set for help-seekers of different severity.

[0334] Step 1204: Set a red mark.

[0335] For example, a red mark may be set for high-risk help seekers.

[0336] Step 1205: Classify the plans.

[0337] Exemplarily, for a help-seeker who is not at a high risk level, the initial plan for the help-seeker can be determined using the plan classification model provided in the aforementioned embodiment.

[0338] Step 1206: Obtain a target question list and target answer data.

[0339] Exemplarily, the method provided in the aforementioned embodiment can be used to obtain the first question list and the second question list corresponding to the initial plan, as well as the answer data of each of the above question lists, so as to obtain the target question list and target answer data.

[0340] For example, the target question list and the target answer data may be further analyzed to determine whether the help-seeker is in a critical or normal condition.

[0341] Exemplarily, if the help-seeker is in a critical condition, step 1207 may be executed, and if the help-seeker is in a normal condition, step 1208 may be executed.

[0342] Step 1207: Set a yellow mark.

[0343] For example, a yellow mark may be set for a person seeking help in a critical condition.

[0344] Step 1208: Set a green mark.

[0345] For example, a green mark may be set for help seekers of average level.

[0346] Through the above process, the embodiment of the present application realizes setting different color identifications for help seekers of different severity levels, thereby realizing intuitive marking of the status of the help seekers.

[0347] Based on the foregoing embodiment, the rescue guidance method provided in the embodiment of the present application may further include the following steps:

[0348] The state change data of the help seeker is obtained, and the parameters of the target network model are adjusted based on the state change data.

[0349] In one embodiment, the state change data may be acquired after a target period of time; illustratively, the target period of time may be determined or adjusted according to the severity, for example, the target period of time may be two days or three days.

[0350] In one embodiment, the state change data may include the change data of the patient state of the help-seeker after the help-seeker performs self-help according to the rescue guidance, or the target object rescues the help-seeker according to the rescue guidance.

[0351] In one implementation, parameters of at least one module in the first model and the second model may be adjusted based on the state change data.

[0352] In one implementation, adjusting the parameters of the target network model may be achieved in the following manner:

[0353] Based on the difference between the state change data and the help-seeking data, the parameters related to the granularity of the rescue guidance categories in the target network model are adjusted, thereby optimizing the precision of the rescue guidance categories determined by the target network model.

[0354] Based on the matching degree between the state change data and the help-seeking data, the parameters related to the accuracy of the rescue guidance category in the target network model are adjusted, thereby optimizing the accuracy of the rescue guidance category determined by the target network model.

[0355] The help seeker's suggestion data or opinion data for the target network model is obtained, and the parameters of the target network model are adjusted in a targeted manner according to the suggestion data or opinion data and the state change data.

[0356] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application, after obtaining the rescuer's state change data, can adjust the parameters of the target network model based on the state change data, thereby realizing closed-loop adjustment and optimization of the target network model. In this way, during the application process of the target network model, through the above-mentioned closed-loop adjustment, the target network model can be continuously optimized in at least one dimension.

[0357] Based on the above embodiments, in the rescue guidance method provided in the embodiments of the present application, the severity of the help-seeking event corresponding to the help-seeker is determined based on the help-seeking data, which can be achieved in the following ways:

[0358] The help request data is classified through the classification model to determine the severity.

[0359] Exemplarily, the classification model may include the critical level classification model in the aforementioned embodiment.

[0360] Accordingly, the rescue guidance method provided in the embodiment of the present application may further include the following steps:

[0361] Obtain the status change data of the person seeking help; and adjust the parameters of the classification model based on the status change data.

[0362] In one implementation, based on the state change data, the network parameters of the plan classification model can also be optimized simultaneously, thereby optimizing the classification performance of the plan classification model.

[0363] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application, after classifying the help-seeking data through the classification model to determine the severity, can also adjust the parameters of the classification model based on the state change data of the help-seeker, thereby realizing continuous and automated closed-loop optimization of the classification model.

[0364] Fig.13 This is a schematic diagram of the process of rescuing a help-seeker provided in the embodiment of the present application. Fig.13 As shown, the process may include the following steps:

[0365] Step 1301: Obtain chief complaint data.

[0366] Step 1302: High-risk classification.

[0367] Exemplarily, the chief complaint data may be processed by the method provided in the aforementioned embodiment to obtain help-seeking data, and the help-seeking data may be classified by the critical level classification model provided in the aforementioned embodiment to obtain a classification result.

[0368] Step 1303: Determine whether it is high risk.

[0369] Exemplarily, if the classification result indicates that the severity of the help-seeker is at a high-risk level, step 1304 may be executed; if the classification result indicates that the severity of the help-seeker is not at a high-risk level, step 1305 may be executed.

[0370] Step 1304: Determine whether there is breathing.

[0371] Exemplarily, if the person seeking help is in a breathing state, the breathing state data can be obtained and step 1306 is executed. If the person seeking help is in a non-breathing state, the breathing state data can be recorded as “non-breathing” and step 1307 is executed.

[0372] Step 1305: Classify the plans.

[0373] Exemplarily, the plan classification can be performed through the plan classification model to obtain an initial plan.

[0374] Exemplarily, after step 1305 is completed, step 1313 can be executed.

[0375] Step 1306: Determine whether the person is conscious.

[0376] Exemplarily, if the person seeking help is in a conscious state, the consciousness state data is obtained and step 1307 is executed. If the person seeking help is in an unconscious state, the consciousness state data can be recorded as “unconscious” and step 1307 is executed.

[0377] Step 1307: Determine the high-risk rescue guidance category.

[0378] Exemplarily, the high-risk rescue guidance category may be the first rescue guidance category in the aforementioned embodiment.

[0379] Exemplarily, the help-seeking data, the respiratory status data, and the consciousness status data may be input into the first model provided in the aforementioned embodiment, and a high-risk rescue guidance category may be obtained, and then step 1308 may be executed.

[0380] Step 1308: Implement rescue guidance.

[0381] Exemplarily, a target rescue guidance strategy corresponding to a high-risk rescue guidance category or a general rescue guidance category may be implemented for the help-seeker.

[0382] After step 1308 is completed, steps 1309 to 1310 may be performed.

[0383] Step 1309: Obtain guidance results and feedback.

[0384] For example, after a specified period of time when the rescue guidance is implemented, the rescuer may be contacted by phone or via the Internet to obtain the guidance results and feedback provided by the rescuer.

[0385] Exemplarily, the guidance result may be the state change data in the aforementioned embodiment.

[0386] For example, the model that needs to be optimized can be determined from the high-risk patient guidance model, the emergency plan classification model, and the high-risk patient classification model according to the severity of the help-seeker who provides guidance results and feedback when initiating the help application.

[0387] Exemplarily, if the severity level is a high risk level, step 1310 and step 1311 may be executed, and if the severity level is not a high risk level, step 1312 may be executed.

[0388] Step 1310: Optimize the high-risk patient guidance model.

[0389] Exemplarily, the high-risk patient guidance model may be the first model in the aforementioned embodiment.

[0390] Step 1311, optimize the high-risk patient classification model.

[0391] Exemplarily, the high-risk patient classification model may be the critical level classification model in the aforementioned embodiment.

[0392] Step 1312: Optimize the emergency plan classification model.

[0393] Step 1313: Obtain the target question sequence.

[0394] Exemplarily, the target question sequence may include a first question list corresponding to the initial plan.

[0395] Exemplarily, if the answer data of the first question list is associated with at least one question outside the first question list, it can be determined that the plan associated with the at least one question outside the first question list is a jump plan, and step 1314 is executed.

[0396] Exemplarily, after determining that the plan no longer jumps, or the plan jump does not occur, step 1315 can be executed.

[0397] Step 1314: Determine whether to jump to the plan.

[0398] Exemplarily, if a plan jump operation is performed, a second question list and corresponding answer data are obtained, and the process switches to step 1313 ; if no plan jump occurs, step 1315 is executed.

[0399] Step 1315: Determine the general rescue guidance category.

[0400] Exemplarily, the general rescue guidance category may be the second rescue guidance category in the aforementioned embodiment.

[0401] Exemplarily, the second model provided in the aforementioned embodiment can be used to classify the target question list, target response data, and help-seeking data to determine the second rescue guidance category.

[0402] Exemplarily, after step 1315 is completed, step 1308 can be executed.

[0403] From the above, it can be seen that the rescue guidance method provided in the embodiment of the present application can not only automatically and intelligently classify the serious conditions of the help-seekers, but also collect data related to the severity of the help-seekers of different categories of severity in a targeted manner, thereby providing a data basis for determining the rescue guidance category for the help-seekers and implementing rescue guidance, thereby improving the accuracy of the rescue guidance category and the rescue guidance strategy, and improving the targetedness and efficiency of the rescue guidance; and after obtaining the guidance results and feedback, it can optimize the high-risk patient classification model, the high-risk patient guidance model, and the plan classification model based on the guidance results and feedback, thereby realizing the closed-loop and continuous optimization and upgrading of the above-mentioned various models in the actual rescue guidance process.

[0404] Fig.14 A schematic diagram of the structure of the rescue guidance system provided in the embodiment of the present application is shown in FIG. Figure 4 As shown, the system may include a client 1401 and a server 1402 .

[0405] The client 1401 may include an electronic device that the help seeker can operate or hold; for example, the electronic device may include: Fig.14 The first user device and the second user device shown in the figure, wherein the first user device may include a smart wearable device, such as a smart watch, etc., through which the blood pressure, blood sugar, blood oxygen, sleep, heart rate and other data of the person seeking help can be collected and recorded in real time; the second user device may include a mobile smart device that can be operated by the person seeking help, such as a notebook computer, etc., through which the audio data, video data, image data and text data of the person seeking help can be collected; illustratively, after the person seeking help sends a request for help to the server 1402, the data collected by the first user device and the second user device respectively can be sent to the server 1402.

[0406] Exemplarily, the server 1402 may store a first aid knowledge base, a critical level model, a plan classification model, a rescue guidance model, and first aid medical record data in its database; wherein, the first aid knowledge base and first aid medical record data are used by the server 1402 to convert and correct various data sent by the client 1401, thereby obtaining help data.

[0407] Exemplarily, the rescue guidance model may include the first model and the second model in the aforementioned embodiments.

[0408] In the actual rescue guidance process, the server 1402 can determine the target rescue guidance strategy for the help-seeker according to the rescue guidance method provided in the aforementioned embodiment, and implement rescue guidance for the help-seeker through the strategy.

[0409] During the actual rescue guidance process, the server 1402 can also output a list of target questions and instructions for obtaining respiratory status and consciousness status to the client 1401. The client 1401 can respond to the above instructions, generate reply data, respiratory status data, and consciousness status data, and send the above data to the server 1402.

[0410] For example, after obtaining the status change data and suggestion data of the person seeking help, the server 1402 can optimize the high-risk classification model, the emergency plan classification model, the rescue guidance model, etc. according to the status change data and the suggestion data.

[0411] From the above, it can be seen that in the embodiment of the present application, the client and the server are organically combined, so that the server can process the rescue request sent by the client in real time, automatically and intelligently, and can also determine the rescue guidance strategy in a targeted and accurate manner, thereby improving the automation and intelligence level of rescue guidance, improving the efficiency of rescue guidance, and improving the effect of rescue guidance.

[0412] Based on the foregoing embodiments, an embodiment of the present application further provides an electronic device, the electronic device comprising a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, a rescue guidance method as provided in any of the previous embodiments can be implemented.

[0413] Based on the foregoing embodiments, an embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor of an electronic device, it can implement the rescue guidance method provided in any of the previous embodiments.

[0414] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0415] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0416] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0417] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0418] It should be noted that the above-mentioned computer-readable storage medium can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM) and other memories; it can also be various electronic devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0419] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0420] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0421] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus necessary general hardware nodes, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0422] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0423] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0424] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0425] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A rescue guidance method, characterized in that: The method comprises: Obtain the help-seeking data of the help-seeker; Based on the help-seeking data, determining the severity of the help-seeking event corresponding to the help-seeker; Based on the severity, determining a target rescue guidance strategy for the help-seeker; Based on the target rescue guidance strategy, the help seeker is guided to handle the help-seeking event.

2. The method according to claim 1, characterized in that Determining a target rescue guidance strategy for the help-seeker based on the severity includes: determining a target network model associated with the severity; Obtain patient status data; Processing the patient status data through the target network model to obtain a target rescue category; Determine the rescue guidance strategy corresponding to the target rescue category as the target rescue guidance strategy.

3. The method according to claim 2, characterized in that If the severity is greater than or equal to the target severity, obtaining patient status data includes: Acquiring the breathing state data and / or consciousness state data of the person seeking help; The respiratory state data and / or the consciousness state data, and the help-seeking data are fused to obtain the patient state data.

4. The method according to claim 3, characterized in that The step of obtaining the respiratory status data of the help seeker includes: Outputting a first prompt message to prompt the user to place a hand on the chest of the help-seeker and collect the ups and downs of the chest where the hand is placed; Based on the fluctuation state, the breathing state data is determined.

5. The method according to claim 3, characterized in that: The step of obtaining the consciousness state data of the help seeker includes: Outputting a second prompt message to prompt the help-seeker to tap at least one body part and collecting the help-seeker's reaction state; Based on the reaction state, the consciousness state data is determined.

6. The method according to claim 2, characterized in that If the severity is greater than or equal to the target severity, the target network model includes a first model; the first model includes a first mapping module, a first association module, a first extraction module and a first classification module; The step of processing the patient status data by the target network model to obtain a target rescue category includes: Mapping the patient status data by the first mapping module to obtain a first mapping result; Performing association processing on the first mapping result by the first association module to obtain a first association result; Performing feature extraction processing on the first association result by the first extraction module to obtain a first extraction result; Classifying the first extraction result by the first classification module to obtain the first guidance category; Determine the first guidance category as the target rescue category.

7. The method according to claim 6, characterized in that The first extraction module includes a forward network module including N extraction units; the N extraction units are respectively connected to the first classification module; N is an integer greater than 3; the first extraction module performs feature extraction processing on the first association result to obtain a first extraction result, including: Performing classification processing on the extraction result output by the n-1th extraction unit in the first extraction module through the first classification model to obtain a first classification probability; wherein n is an integer greater than 1 and less than N; Performing feature extraction on the first classification probability and the extraction result of the n-1th extraction unit through the nth extraction unit to obtain the extraction result of the nth extraction unit; The classifying the first extraction result by the first classification module to obtain the first guidance category includes: Processing the extraction results of the first extraction unit to the extraction results of the Nth extraction unit by the first classification module to obtain N classification probabilities; Based on the N classification probabilities, the first guidance category is determined.

8. The method according to claim 6, characterized in that The first model is obtained by training the first model in the initial state based on sample data; the sample data includes a high-risk rescue guidance mark; the method further includes: Mapping the sample data by the first mapping module to obtain first vector data; Performing context association on the first vector data by the first association module to obtain first context data; Processing the high-risk rescue guidance identifier and the first context data through the first extraction module and the first classification module to obtain a prediction identifier; Based on the matching degree between the prediction mark and the high-risk rescue guidance mark, at least the weight parameter of the first extraction module is adjusted to obtain the first model.

9. The method according to claim 8, characterized in that The first extraction module includes a forward network module including N extraction units; the N extraction units are respectively connected to the first classification module; N is an integer greater than or equal to 3; the high-risk rescue guidance identifier and the first context data are processed by the first extraction module and the first classification module to obtain a prediction identifier, including: The n-1th high-risk rescue guidance identifier is weighted with a first weight to obtain a first result; wherein the n-1th high-risk rescue guidance identifier includes the high-risk rescue guidance identifier of the n-1th data in the sample data; n is an integer greater than 1 and less than N; The n-1th guiding category probability output by the first classification module is weighted with a second weight to obtain a second result; wherein the n-1th guiding category probability is at least associated with the output data of the n-1th extraction unit in the forward network module; and the output data of the n-1th extraction unit is associated with the nth data in the sample data; Processing the output data of the n-1th extraction unit, the first result, and the second result through the nth extraction unit in the forward network module to obtain the output data of the nth extraction unit; The output data from the first extraction unit to the output data of the Nth extraction unit are classified by the first classification module to obtain the prediction identifier.

10. The method according to claim 8, characterized in that Mapping the sample data by the first mapping module to obtain first vector data includes: Counting the high-risk rescue guidance identifiers to obtain the m-th identifier number corresponding to the high-risk rescue guidance identifier of the m-th guidance category; wherein m is an integer greater than or equal to 1 and less than or equal to M; M is the number of rescue guidance categories included in the high-risk rescue guidance identifier; Arrange the first number of identifiers to the Mth number of identifiers in descending order to obtain a sorting result; Arrange the data associated with the first guidance category to the data associated with the Mth guidance category in the sample data based on the ranking result degree to obtain an arrangement result; The arrangement result is mapped by the first mapping module to obtain the first vector data.

11. The method according to claim 2, characterized in that If the severity is less than the target severity, obtaining patient status data includes: Obtain a list of target questions including patient status questions; Obtaining target answer data of the help seeker to the questions in the target question list; Based on the target question list and the target answer data, the patient status data is determined.

12. The method according to claim 11, characterized in that The target question list includes a first question list associated with the initial plan; the step of obtaining the target question list including patient status questions includes: Classifying the help-seeking data through a plan classification model to obtain the initial plan; A first list of questions associated with the initial plan is obtained.

13. The method according to claim 11, characterized in that The target question list includes a second question list; and obtaining the target question list including patient status questions includes: Classify the help-seeking data through a plan classification model to obtain an initial plan; Obtaining answer data for a first list of questions associated with the initial plan category; If the answer data is associated with at least one question outside the first question list, determining that the plan associated with the at least one question is a jump plan; Determine that the question list associated with the jump plan is the second question list.

14. The method according to claim 11, characterized in that The target network model includes a second model; the second model includes a second mapping module, a second association module, a second extraction module and a second classification module; the patient status data is processed by the target network model to obtain a target rescue category, including: Obtain the target plan associated with the target problem list; Processing the plan name of the target plan, the request-seeking data, the target question list, and the target answer data through the second mapping module to obtain a second mapping result; Performing context association on the second mapping result by the second association module to obtain a second association result; Processing the second association result by the second extraction module and the second classification module to obtain the second guidance category; Determine the second guidance category as the target rescue category.

15. The method according to claim 2, characterized in that The method further comprises: Acquiring status change data of the help seeker; Based on the state change data, parameters of the target network model are adjusted.

16. The method according to claim 1, characterized in that The determining, based on the help-seeking data, the severity of the help-seeking event corresponding to the help-seeker includes: Classify the help-seeking data by using a classification model to determine the severity; The method further comprises: Acquiring status change data of the help seeker; Based on the state change data, parameters of the classification model are adjusted.

17. An electronic device, characterized in that: The electronic device includes a processor and a memory; a computer program is stored in the memory; when the computer program is executed by the processor, it can implement the rescue guidance method as described in any one of claims 1 to 16.

18. A computer-readable storage medium, characterized in that: The storage medium stores a computer program; when the computer program is executed by a processor of an electronic device, it can implement the rescue guidance method as described in any one of claims 1 to 16.

Images