Information processing method and device, computer device and storage medium

Abstract

The application relates to an information processing method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining abnormal description information specified by a target object in a current dialogue round; determining an abnormal category probability distribution corresponding to the abnormal description information based on a Bayesian network between an abnormal category and a symptom category, and determining a first symptom category probability distribution according to an abnormal symptom transfer matrix and the abnormal category probability distribution; determining a candidate symptom category according to the first symptom category probability distribution, entering a next dialogue round, and obtaining a supplementary symptom category selected by the target object from the candidate symptom category; updating the abnormal description information based on the supplementary symptom category, returning to the step of determining the joint probability distribution based on the abnormal category and the symptom category, and continuing to execute until a preset stop condition is reached; and outputting corresponding recommendation information to the target object based on the finally updated abnormal description information. The method can improve the efficiency of generating recommendation information.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an information processing method, apparatus, computer device, and storage medium. Background Art

[0002] With the development of internet technology, intelligent medical consultation systems have emerged, enabling online consultations. Even with these systems, patients often don't know which department should be consulted for their symptoms, nor which online doctor to seek treatment. Consequently, they rely on intuition to choose a department and doctor within the system.

[0003] Currently, patients typically search online for questions like "Which department should I go to for a headache?" or "Which doctor treats headaches?" They then manually filter the resulting webpages to identify the appropriate department or doctor, and then proceed with online consultations based on these recommendations. This reliance on manual screening to identify symptom-related recommendations is inefficient. Summary of the Invention

[0004] Based on this, it is necessary to provide an information processing method, device, computer equipment and storage medium that can improve the efficiency of recommendation information generation in response to the above technical problems.

[0005] An information processing method, the method comprising:

[0006] Get the exception description information specified by the target object in the current dialogue round;

[0007] Determine the abnormality category probability distribution corresponding to the abnormality description information based on a Bayesian network between the abnormality category and the symptom category, and determine the first symptom category probability distribution according to the abnormal symptom transfer matrix and the abnormality category probability distribution;

[0008] Determining candidate symptom categories based on the first symptom category probability distribution, and entering the next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories;

[0009] Updating the abnormality description information based on the supplementary symptom category, returning the step of determining the abnormality category probability distribution corresponding to the abnormality description information based on the joint probability distribution between the abnormality category and the symptom category, and continuing to execute until a preset stopping condition is reached;

[0010] Based on the finally updated abnormal description information, corresponding recommendation information is output to the target object.

[0011] In one embodiment, obtaining the exception description information specified by the target object in the current conversation turn includes:

[0012] Display a dialogue interface; the dialogue interface includes an information input box and a symptom category display area;

[0013] If the current dialogue round is the first dialogue round, in response to the information input operation triggered by the information input box, obtaining abnormal description information of the first dialogue round;

[0014] If the current conversation round is not the first conversation round, then in response to the selection operation of the candidate symptom category in the symptom category display area, the selected supplementary symptom category is obtained, and the abnormal description information of the current conversation round is updated based on the supplementary symptom category to obtain the abnormal description information of the next conversation round.

[0015] In one embodiment, if the current conversation turn is not the first conversation turn, in response to a selection operation on a candidate symptom category in the symptom category display area, a selected supplementary symptom category is obtained, including:

[0016] For each dialogue round except the first dialogue round, the candidate symptom categories determined in the previous dialogue round are displayed in the symptom category display area of ​​the current dialogue round;

[0017] In response to a selection operation on a candidate symptom category in the symptom category display area, a selected supplementary symptom category is obtained.

[0018] An information processing device, comprising:

[0019] The information acquisition module is used to obtain the abnormal description information specified by the target object in the current dialogue round;

[0020] a supplementary symptom determination module, configured to determine, based on a Bayesian network between abnormality categories and symptom categories, a probability distribution of abnormality categories corresponding to the abnormality description information, and determine a first symptom category probability distribution based on an abnormality symptom transfer matrix and the abnormality category probability distribution; determine a candidate symptom category based on the first symptom category probability distribution, enter a next dialogue round, and obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the joint probability distribution based on the abnormality category and the symptom category, and continue to determine the abnormality category probability distribution corresponding to the abnormality description information until a preset stopping condition is reached;

[0021] The recommendation information generation module is used to output corresponding recommendation information to the target object based on the abnormal description information finally updated.

[0022] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any information processing method provided in the embodiments of the present application when executing the computer program.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any information processing method provided in the embodiments of the present application.

[0024] A computer program product or computer program includes computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any information processing method provided in the embodiments of the present application.

[0025] The above-mentioned information processing method, apparatus, computer equipment, storage medium and computer program can determine the abnormal category probability distribution corresponding to the abnormal description information based on the Bayesian network by obtaining the abnormal description information specified by the Bayesian network and the target object in the current dialogue round, and thus can determine the first symptom category probability distribution corresponding to the abnormal category probability distribution based on the abnormal symptom transfer matrix. By obtaining the first symptom category probability distribution, the candidate symptom category to be inquired in the next step can be accurately predicted based on the first symptom category probability distribution, thereby improving the accuracy of the candidate symptom category. By determining the candidate symptom category to be inquired in the next step, the supplementary symptom category selected by the target object from the candidate symptom category can be obtained. In this way, the abnormal description information can be updated based on the supplementary symptom category, and the step of determining the abnormal category probability distribution corresponding to the abnormal description information based on the joint probability distribution between the abnormal category and the symptom category is returned and continued until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommended information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommended information based on the continuously updated abnormal description information, the efficiency of generating recommended information is greatly improved compared to manual screening.

[0026] In addition, since the present application prioritizes determining the probability value of the target object corresponding to each abnormal category when it has the current abnormal description information, and then obtains the probability value of the target object corresponding to each abnormal category when it has the current abnormal description information through the probability value of the target object corresponding to each abnormal category, compared with the traditional prediction of symptom category through symptom category, the prediction logic of the present application based on the probability value of each abnormal category to the probability value of each symptom category is more explainable, so that the probability distribution of the first symptom category determined based on this prediction logic can be more accurate.

[0027] An information processing method, the method comprising:

[0028] Get the exception description information specified by the target object in the current dialogue round;

[0029] Based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on information features of the standard description text, a second symptom category probability distribution is determined;

[0030] determining a candidate symptom category based on the second symptom category probability distribution, and entering a next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories;

[0031] updating the abnormality description information based on the supplementary symptom category, returning the abnormality description information, generating a standard description text with a fixed symptom order, and continuing the step of determining a second symptom probability distribution based on information features of the standard description text until a preset stopping condition is reached;

[0032] Based on the exception description information updated in the last dialogue round, corresponding recommendation information is output to the target object.

[0033] An information processing device, comprising:

[0034] The exception determination module is used to obtain the exception description information specified by the target object in the current dialogue turn;

[0035] an information updating module configured to generate a standard description text having a fixed symptom order based on the abnormal description information, and determine a second symptom category probability distribution based on information features of the standard description text; determine a candidate symptom category based on the second symptom category probability distribution, and enter a next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormal description information based on the supplementary symptom category, and return the step of generating a standard description text having a fixed symptom order based on the abnormal description information and determining a second symptom probability distribution based on information features of the standard description text, and continue executing until a preset stopping condition is reached;

[0036] The information output module is used to output corresponding recommendation information to the target object based on the abnormal description information updated in the last dialogue round.

[0037] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any information processing method provided in the embodiments of the present application when executing the computer program.

[0038] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any information processing method provided in the embodiments of the present application.

[0039] A computer program product or computer program includes computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any information processing method provided in the embodiments of the present application.

[0040] The above-mentioned information processing method, device, computer equipment, storage medium and computer program can generate a standard description text with a fixed symptom order corresponding to the abnormal description information by obtaining the abnormal description information specified by the target object in the current dialogue round, so that more accurate information features can be extracted from the standard description text with a fixed symptom order, and a more accurate second symptom category probability distribution can be generated based on the more accurate information features. By generating the second symptom category probability distribution, the candidate symptom category to be inquired in the next step can be predicted based on the second symptom category probability distribution, so that the supplementary symptom category selected by the target object can be obtained through the candidate symptom category to be inquired in the next step. In this way, the abnormal description information can be updated based on the supplementary symptom category, and the step of generating a standard description text with a fixed symptom order based on the abnormal description information and determining the second symptom probability distribution based on the information features of the standard description text is continued until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommended information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommended information based on the continuously updated abnormal description information, the efficiency of generating the recommended information is greatly improved compared to the traditional manual screening.

[0041] In addition, since the text processing process is very sensitive to word order, sorting the extracted symptom categories and symptom states to obtain a standard description text with a fixed symptom order can make the extracted information features more accurate, thereby making the probability distribution of the second symptom category determined based on the information features more accurate.

[0042] An information processing method, the method comprising:

[0043] Get the exception description information specified by the target object in the current dialogue round;

[0044] Determining an abnormality category probability distribution based on the abnormality description information, and determining a related first symptom category probability distribution based on the abnormality category probability distribution;

[0045] Based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on information features of the standard description text, a second symptom category probability distribution is determined;

[0046] Determining candidate symptom categories based on the first symptom category probability distribution and the second symptom category probability distribution, and entering the next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories;

[0047] Updating the abnormality description information based on the supplementary symptom category, returning the step of determining the abnormality category probability distribution corresponding to the abnormality description information based on the joint probability distribution between the abnormality category and the symptom category, and continuing to execute until a preset stopping condition is reached;

[0048] Based on the exception description information updated in the last dialogue round, corresponding recommendation information is output to the target object.

[0049] An information processing device, characterized in that the device comprises:

[0050] The information specification module is used to obtain the abnormal description information specified by the target object in the current dialogue round;

[0051] An information adjustment module is configured to determine an abnormality category probability distribution based on the abnormality description information, and determine a related first symptom category probability distribution based on the abnormality category probability distribution; generate a standard description text with a fixed symptom order based on the abnormality description information, and determine a second symptom category probability distribution based on information features of the standard description text; determine a candidate symptom category based on the first symptom category probability distribution and the second symptom category probability distribution, and enter a next dialogue round to obtain a supplementary symptom category selected by the target object from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the joint probability distribution based on the abnormality category and the symptom category, and continue to perform the step of determining the abnormality category probability distribution corresponding to the abnormality description information until a preset stopping condition is reached;

[0052] The output display module is used to output corresponding recommendation information to the target object based on the abnormal description information updated in the last dialogue round.

[0053] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any information processing method provided in the embodiments of the present application when executing the computer program.

[0054] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any information processing method provided in the embodiments of the present application.

[0055] A computer program product or computer program includes computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any information processing method provided in the embodiments of the present application.

[0056] The above-mentioned information processing method, apparatus, computer equipment, storage medium and computer program can generate a first symptom category probability distribution and a second symptom category probability distribution corresponding to the abnormal description information by obtaining the abnormal description information specified by the target object in the current dialogue round, thereby accurately predicting the candidate symptom category that needs to be inquired about next based on the first symptom category probability distribution and the second symptom category probability distribution. By accurately predicting the candidate symptom category that needs to be inquired about next, the supplementary symptom category selected by the target object from the candidate symptom category can be obtained. In this way, the abnormal description information can be updated based on the supplementary symptom category until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommendation information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommendation information based on the continuously updated abnormal description information, the efficiency of generating recommendation information is greatly improved compared to traditional manual screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A diagram of an application environment of an information processing method in one embodiment;

[0058] Figure 2 This is a schematic diagram of a dialogue interface in one embodiment;

[0059] Figure 3 1 is a flow chart of an information processing method in one embodiment;

[0060] Figure 4 Schematic diagram of a Bayesian network between abnormality categories and symptom categories in one embodiment;

[0061] Figure 5 is a schematic diagram of an abnormal symptom transfer matrix in one embodiment;

[0062] Figure 6 is a schematic diagram of a dialogue interface in another embodiment;

[0063] Figure 7 is a schematic diagram of an initial probability distribution in one embodiment;

[0064] Figure 8 1 is a flow chart of an information processing method in one embodiment;

[0065] Figure 9 is a flowchart of an information processing method in another embodiment;

[0066] Figure 10 is a flowchart of an information processing method in a specific embodiment;

[0067] Figure 11 is a structural block diagram of an information processing device in one embodiment;

[0068] Figure 12 is a structural block diagram of an information processing device in another embodiment;

[0069] Figure 13 is a structural block diagram of an information processing device in yet another embodiment;

[0070] Figure 14 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0072] Figure 1 FIG1 is an application environment diagram describing an information processing method in an embodiment. Figure 1 The information processing method is applied to an information processing system 100. The information processing system 100 includes a terminal 102 and a server 104. The terminal 102 can be used alone to execute the information processing method of the present application, or the terminal 102 and the server 104 can be used together to execute the information processing method of the present application. Taking the example of the terminal 102 and the server 104 working together to execute the information processing method of the present application, the terminal 102 can run a medical communication application, through which multiple rounds of conversations can be conducted with the user, and corresponding recommendation information can be output based on the multiple rounds of conversations, such as recommended departments or recommended doctors. In each round of the multiple rounds of conversations, the terminal 102 can collect the symptom type entered by the user in the current conversation round, obtain abnormality description information for the current conversation round, and send the abnormality description information for the current conversation round to the server 104, so that the server 104 can output candidate symptom types based on the received abnormality description information and return the candidate symptom types to the terminal 102. Terminal 102 can proceed to the next round of conversation and display the received candidate symptom types, allowing the user to select a supplementary symptom type from the displayed candidate symptom types. Terminal 102 can then update the abnormality description information determined in the current conversation round based on the supplementary symptom type, obtaining the abnormality description information for the next conversation round. This process continues in this manner until the abnormality description information for the final conversation round is obtained. Server 104 receives corresponding push information based on the final updated abnormality description information and returns the push information to Terminal 102 for display.

[0073] Among them, the server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal 102 can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a car terminal, a smart TV, etc., but is not limited to these. The terminal 102 and the server 104 can be directly or indirectly connected via wired or wireless communication, and this application does not limit this. The medical communication application is a communication application suitable for the medical field, which can be an application client or a sub-application running in a parent application, etc., and this embodiment of the application does not limit this. It should also be noted that other applications can also be run on the terminal, which can provide communication functions, thereby realizing communication dialogues with the server, and this embodiment of the application does not limit this. For example, if a browser application is running on the terminal, the user can use the browser to start a dialogue and enter the channel for communicating with the server in the browser application, thereby realizing multiple dialogue turns.

[0074] In order to better understand the information processing method in the embodiments of the present application, the overall application scenario of the present application is introduced below:

[0075] For the medical communication system, its most critical function is to ask the user about the symptom categories he has, and make corresponding information recommendations for the user based on the symptom categories that have been asked, such as recommending departments and doctors to the user. Since it is difficult for users to input all the symptom categories at one time, most users will only input the most critical symptom categories in the medical communication system, such as inputting the symptom category of "body temperature higher than 38.5 degrees Celsius", and ignore the rest of the non-critical symptom categories, such as ignoring the symptom category of "coughing up phlegm". In response to the above-mentioned defects, the medical communication system of the present invention can guide users to input complete symptom categories through multiple rounds of dialogue, thereby outputting more accurate recommendation information through complete symptom categories. For example, refer to Figure 2 In a medical communication application, a user can enter their chief complaint, specifically their primary symptom category, such as "Headache, I've had it for a while." 201 Based on the symptom category entered by the user, the medical communication application predicts the next symptom category to be inquired about and provides feedback to the user, allowing the user to independently select whether these symptom categories are available. A medical communication application is a communication application in the medical field. Furthermore, the medical communication application predicts the next symptom category to be inquired about based on the symptom category selected by the user. This process continues iteratively until a stopping condition is reached, resulting in a complete set of symptom categories.

[0076] For example, based on the query "I've had a headache for a while" 201, the medical communication application might predict that the next symptom category to be inquired about is "frontal headache," "occipital headache," "bilateral headache," etc., and present the symptom categories to be inquired about via a multiple-selection list 202, guiding the user to select the headache category they have. If the user selects "unilateral headache," the medical communication application again predicts that the next symptom category to be inquired about is "dizziness," based on the query "I've had a headache for a while" and "unilateral headache," allowing the user to select whether they have dizziness. Finally, based on the complete set of symptom categories inquired about during multiple rounds of conversation, the medical communication application outputs recommended departments, recommended doctors, and other information. Figure 2 A schematic diagram of a conversation interface of a medical communication application in one embodiment is shown.

[0077] In one embodiment, Figure 3 As shown, an information processing method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, and the computer device can be the above Figure 1 The terminal 102 or server 104 in the information processing method includes the following steps:

[0078] Step S302: Obtain the exception description information specified by the target object in the current dialogue round.

[0079] A conversation round refers to a conversation round between a medical communication application and a target object. A complete conversation may include multiple rounds of conversations. Through multiple rounds of conversations, the medical communication application can understand at least one symptom category that the target object currently has. A conversation round refers to the process in which the target object initiates a conversation and the medical communication application responds to the conversation. For example, refer to Figure 2 The process of the target subject selecting "unilateral headache" and the medical communication application displaying "Do you have the following symptoms: 'dizziness', 'none of the above'?" is called a conversation turn. Within a conversation turn, the target subject can initiate the current conversation turn by selecting a supplementary symptom category from the candidate symptom categories output in the previous turn. The medical communication application can then output a symptom category requiring further inquiry based on the supplementary symptom category selected by the target subject and use the outputted symptom category as the response in the current conversation turn.

[0080] Abnormal description information refers to the physiological abnormality information of the target object, and the abnormal description information may specifically include symptom categories and symptom states. For example, the abnormal description information may specifically be "the target object has a headache, a headache on one side, and dizziness". Each symptom category may include at least two symptom states. For example, when the symptom category is "dizziness", the corresponding symptom state may be "yes" or "no". When the symptom category is "dizziness" and the corresponding symptom state is "yes", it can be determined that the target object has dizziness, thereby obtaining the abnormal description information "the target object has a headache". It is easy to understand that since a new symptom category will be obtained in each dialogue round, the abnormal description information of the target object can be continuously updated with multiple rounds of dialogue. For example, in the initial dialogue round, the abnormal description information of the target object may be "headache". In the second round of dialogue, the abnormal description information may be updated to "the target object has a headache and dizziness on one side" based on the symptom category supplemented by the target object.

[0081] Specifically, for the current conversation round of multiple conversation rounds, the medical communication application in the terminal can determine the abnormal description information of the target object in the current conversation round. Among them, for the first conversation round of multiple conversation rounds, the medical communication application can use the symptom category complained by the target object as the abnormal description information in the initial conversation round; for the remaining conversation rounds except the first conversation round of multiple conversation rounds, the medical communication application can use the abnormal description information in the current conversation round and the supplementary symptom category selected in the current conversation round as the abnormal description information of the next conversation round. For example, if the abnormal description information of the current conversation round is "the user has a headache" and the symptom category inquiry list displayed by the current medical communication application is Figure 2 202 in , and when the symptom category selected by the target object from 202 is "unilateral headache", the abnormal description information of the current dialogue round can be updated to "the user has a headache, a headache on one side".

[0082] In one embodiment, when the target subject selects a supplementary symptom category from the candidate symptom categories, the medical communication application may determine the symptom status corresponding to the supplementary symptom category, thereby updating the abnormality description information based on the supplementary symptom category and the corresponding symptom status. For example, the medical communication application may default the symptom status of the supplementary symptom category selected by the target subject, except for "not above", to "yes". Figure 2 When the target object selects "dizziness", the medical communication application may default the symptom status of "dizziness" to "yes". When the target object selects "none of the above", the medical communication application may default the symptom status of "dizziness" to "no".

[0083] In one embodiment, the target subject may input the symptom category of the main complaint by voice input or text input, so that the medical communication application may use the symptom category of the main complaint of the target subject as the abnormality description information in the first dialogue round.

[0084] In one embodiment, obtaining the abnormal description information specified by the target object in the current dialogue round includes: displaying a dialogue interface; the dialogue interface includes an information input box and a symptom category display area; if the current dialogue round is the first dialogue round, then in response to the information input operation triggered by the information input box, the abnormal description information of the first dialogue round is obtained; if the current dialogue round is not the first dialogue round, then in response to the selection operation of the candidate symptom category in the symptom category display area, the selected supplementary symptom category is obtained, and the abnormal description information of the previous dialogue round is updated based on the supplementary symptom category to obtain the abnormal description information of the current dialogue round.

[0085] Among them, the dialogue interface refers to the interface displayed by the medical communication application for communicating with the target object. The dialogue interface may display an information input box and / or a symptom category display area. The abnormal description information of the first dialogue round can be obtained through the information input box, and the abnormal description information of the remaining dialogue rounds except the first dialogue round can be obtained through the symptom category display area. Candidate symptom categories refer to symptom categories that need to be inquired about. The medical communication application can predict the symptom category that needs to be inquired about next through the abnormal description information. Supplementary symptom categories refer to symptom categories supplemented by the target object. The target object can select the corresponding supplementary symptom category from the candidate symptom categories. The previous dialogue round refers to a dialogue round that is located before the current dialogue round and adjacent to the current dialogue round.

[0086] Specifically, a medical communication application may be running on the terminal. When a target user desires to obtain recommended information, the target user may open the medical communication application, which may then display a conversation interface. Furthermore, the target user may enter at least one symptom category they possess through the conversation interface, and the medical communication application may use the symptom category entered by the target user as the abnormality description information for the first conversation turn. Furthermore, when the current conversation turn is a conversation turn other than the first, the conversation interface may display a symptom category display area, displaying candidate symptom categories through the symptom category display area. The target user may then select a supplementary symptom category from the displayed candidate symptom categories, causing the medical communication application to update the abnormality description information for the previous conversation turn based on the supplementary symptom category, thereby obtaining abnormality description information for the current conversation turn. For example, when the current conversation turn is the second conversation turn, the medical communication application may update the abnormality description information for the first conversation turn based on the supplementary symptom category selected by the user, thereby obtaining abnormality description information for the current conversation turn. The medical communication application may then obtain the corresponding candidate symptom category based on the abnormality description information for the current conversation turn.

[0087] In one embodiment, when a medical communication application is started, only an information input box may be displayed in the dialogue interface. When the target object enters the symptom category through the information input box, the dialogue interface may cancel the information input box and switch to displaying the symptom category display area.

[0088] In one embodiment, the medical communication application has a text input control or a voice input control, so that the target object can input the abnormality description information as the first dialogue turn through text input or voice input.

[0089] In the above embodiment, by displaying the information input box and the symptom category display area, the abnormal description information can be determined in time through the displayed information input box and the symptom category display area, thereby improving the efficiency of determining the abnormal description information.

[0090] Step S304: determining the abnormality category probability distribution corresponding to the abnormality description information based on the Bayesian network between the abnormality category and the symptom category, and determining the first symptom category probability distribution according to the abnormal symptom transfer matrix and the abnormality category probability distribution.

[0091] Among them, Bayesian network refers to a probability graphical model, also known as belief network or directed acyclic graph model. Bayesian network consists of nodes, directed edges and parameters. Nodes represent random variables, which include parent nodes and child nodes. Parent nodes can point to child nodes through directed edges. The parameters in the Bayesian network are the joint probability distribution between parent nodes and connected child nodes. Figure 4 , Figure 4The following shows a Bayesian network between abnormality categories and symptom categories in one embodiment. The parent node is the abnormality category, such as "rhinitis", "upper respiratory tract infection", etc. The child node is the symptom category, such as "allergy", "cough", "sore throat", etc. One of the parameters in the Bayesian network can be a joint probability distribution between "rhinitis" and the connected "allergy", "cough", "sore throat" and "dyspnea", that is, one of the parameters in the Bayesian network can be P("rhinitis", "allergy", "cough", "sore throat", "dyspnea"). The joint probability is the probability that multiple random variables meet their respective conditions in a multivariate probability distribution.

[0092] The abnormality category probability distribution corresponding to the abnormality description information refers to the probability value of the target object corresponding to each abnormality category when the target object has the symptom category and symptom status in the abnormality description information. For example, if the abnormality categories are "rhinitis," "pneumonia," and "upper respiratory tract infection," and the current abnormality description information is "the user has allergies and a sore throat," the corresponding abnormality category probability distribution can be: if the target object has allergies and a sore throat, the probability value of the target object corresponding to rhinitis, the probability value of pneumonia, and the probability value of an upper respiratory tract infection.

[0093] The abnormal symptom transfer matrix refers to a matrix used to determine the symptom category based on the abnormal category. Different matrix rows in the abnormal symptom transfer matrix correspond to different abnormal categories, and different matrix columns correspond to different symptom categories. Each element in the abnormal symptom transfer matrix represents the probability value of the corresponding symptom category under the condition that the corresponding abnormal category appears. For example, refer to Figure 5 When the third row in the abnormal symptom transfer matrix corresponds to "abnormal category 2" and the second column corresponds to "symptom category 1", S(3,2)=0.3 means that when the target object has abnormal category 2, the probability value of symptom category 1 is 0.3. Figure 5 A schematic diagram of an abnormal symptom transfer matrix in one embodiment is shown.

[0094] The first symptom category probability distribution includes the probability values ​​corresponding to each symptom category, wherein the first symptom category probability distribution in the current dialogue round refers to the probability value of the target object having each symptom category when the target object corresponds to the abnormal description information of the current dialogue round. For example, when the current abnormal description information is "the user has allergies and sore throat", and the symptom categories are "cough", "dyspnea", "high temperature", and "general fatigue", the first symptom category probability distribution can be: when the target object has allergies and sore throat, the probability value of the target object having a cough, the probability value of having dyspnea, the probability value of having a high temperature, and the probability value of having general fatigue.

[0095] Specifically, upon obtaining the abnormality description information for the current round, the medical communication application may obtain a Bayesian network between the abnormality category and the symptom category, input the abnormality description information into the Bayesian network, and output the abnormality category probability distribution corresponding to the abnormality description information through the Bayesian network. Furthermore, the medical communication application may obtain a pre-set abnormality symptom transition matrix and, based on the abnormality symptom transition matrix and the abnormality category probability distribution, obtain a first symptom category probability distribution. As will be readily understood, the medical communication application may also transmit the obtained abnormality description information for the current round to the server, so that the server can obtain the first symptom category probability distribution based on the received abnormality description information.

[0096] In one embodiment, the first symptom category probability distribution is determined based on the abnormal symptom transfer matrix and the abnormal category probability distribution, including: fusing the abnormal category probability distribution with each matrix column in the abnormal symptom transfer matrix respectively to obtain the first symptom category probability distribution of the current dialogue round.

[0097] Specifically, the medical communication application can fuse the abnormality category probability distribution with each matrix column in the abnormality symptom transfer matrix to obtain the first symptom category probability distribution of the current conversation turn. For example, when the number of abnormality categories is d, the number of symptom categories is s, the abnormality category probability distribution is P_d, and the abnormality symptom transfer matrix is ​​M, the first symptom category probability distribution P_s1 can be M*P_d. Among them, the size of P_d is 1*d, that is, P_d is a matrix with one row and d columns, the size of M is d*s, that is, M is a matrix with d rows and s columns, and the size of P_s1 is 1*s, that is, P_s1 is a matrix with one row and s columns.

[0098] In this embodiment, it is only necessary to directly fuse the abnormal category probability distribution with each matrix column in the abnormal symptom transfer matrix to obtain the first symptom category probability distribution of the current dialogue round, thereby improving the efficiency of determining the first symptom category probability distribution.

[0099] Step S306 , determining candidate symptom categories based on the first symptom category probability distribution, and entering the next dialogue round to obtain the supplementary symptom category selected by the target object from the candidate symptom categories.

[0100] Specifically, since the first symptom category probability distribution contains a probability value corresponding to each symptom category, when the first symptom category probability distribution is obtained, the medical communication application can determine the symptom category in the first symptom category probability distribution that meets the high probability condition, and use the symptom category that meets the high probability condition as a candidate symptom category. The high probability condition can be freely set according to needs. For example, the high probability condition can be that the probability value is greater than or equal to a preset probability threshold, or the high probability condition can be that the probability value has the maximum probability value. It is easy to understand that the server can also determine the candidate symptom category based on the first symptom category probability distribution.

[0101] Furthermore, the medical communication application displays the determined candidate symptom categories and enters the next conversation round, so that the target object can select a supplementary symptom category from the displayed candidate symptom categories.

[0102] In one embodiment, the medical communication application may filter out symptom categories with probability values ​​greater than or equal to a preset probability threshold from the first symptom category probability distribution, and use the symptom categories with probability values ​​greater than or equal to the preset probability threshold as supplementary symptom categories. For example, if the first symptom category probability distribution is: when the target subject has allergies and a sore throat, the probability of the target subject having a cough is 0.4, the probability of having difficulty breathing is 0.3, the probability of having a high temperature is 0.1, and the probability of having general fatigue is 0.2. When the preset probability threshold is 0.3, the candidate symptom categories are cough and difficulty breathing.

[0103] Step S308, updating the abnormality description information based on the supplementary symptom category, returning to the step of determining the abnormality category probability distribution corresponding to the abnormality description information based on the joint probability distribution between the abnormality category and the symptom category, and continuing to execute until a preset stop condition is reached.

[0104] Specifically, when the medical communication application displays candidate symptom categories, the next conversation round begins. At this point, the target user can select a supplementary symptom category from the displayed candidate symptom categories. The medical communication application can then update the abnormality description information in the current conversation round based on the selected supplementary symptom category, obtaining the abnormality description information for the next conversation round. For example, if the current conversation round is M and the next conversation round is M+1, the target user can select a supplementary symptom category from the candidate symptom categories displayed in round M+1. The medical communication application can then update the abnormality description information in round M based on the selected supplementary symptom category, obtaining the abnormality description information for round M+1.

[0105] In the next conversation round, the medical communication application determines the probability distribution of the abnormality category corresponding to the abnormality description information of the next conversation round based on the Bayesian network between the abnormality category and the symptom category, and determines the probability distribution of the first symptom category of the next conversation round based on the abnormal symptom transition matrix and the probability distribution of the abnormality category corresponding to the abnormality description information of the next conversation round. Thus, based on the probability distribution of the first symptom category of the next conversation round, the candidate symptom category of the next conversation round is determined. The medical communication application displays the candidate symptom category of the next conversation round to enter the next conversation round. And so on, until the preset stop condition is reached. The preset stop condition can be freely set according to needs. For example, the medical communication application can stop when a preset conversation round is reached, or when a preset number of symptom categories are obtained. This embodiment is not limited here. The prediction logic of this application based on the probability value of each abnormality category to the probability value of each symptom category is more in line with the logical thinking of natural people, making the prediction logic more interpretable.

[0106] Step S310: outputting corresponding recommendation information to the target object based on the finally updated abnormal description information.

[0107] Specifically, upon obtaining the final updated exception description, the medical communication application can generate and display corresponding recommendations based on the final updated exception description. Recommendations refer to information that pushes relevant content to the target audience. For example, recommendations could include recommended departments, recommended doctors, and recommended diets.

[0108] In one embodiment, the medical communication application stores the correspondence between the abnormal description information and the recommended information. When the final updated abnormal description information is obtained, the medical communication application can determine the recommended information that needs to be pushed to the target object based on the correspondence between the abnormal description information and the recommended information.

[0109] In one embodiment, the medical communication application stores the correspondence between abnormality categories and recommended information. Upon obtaining the final updated abnormality description information, the medical communication application may determine the abnormality category probability distribution corresponding to the final updated abnormality description information based on a Bayesian network between the abnormality category and the symptom category, and extract the target abnormality category with the maximum probability value from the determined abnormality category probability distribution. Furthermore, based on the correspondence between the abnormality category and the recommended information, the medical communication application determines the target recommended information corresponding to the target abnormality category and displays the target recommended information accordingly.

[0110] In one embodiment, when the final updated abnormality description information is obtained, the medical communication application may input the final updated abnormality description information into a pre-trained machine learning model, and output corresponding recommendation information through the pre-trained machine learning model.

[0111] In one embodiment, the conversation interface in the medical communication application may display the conversation content of each round, for example, Figure 6 The medical communication application can display the conversation content of the nth round "Do you have the following symptoms - body temperature greater than 38.5 degrees" 601, the conversation content of the n+1th round "Do you have the following symptoms - cough, sore throat" 602, and the conversation content of the n+1th round "Do you have the following symptoms - no sputum" 603. Figure 6 A schematic diagram of a dialogue interface in one embodiment is shown.

[0112] In one embodiment, referring to Figure 6, when in the final round of dialogue, the target object can choose to continue the intelligent consultation by touching the "Continue" control 604, or can return to the main interface of the medical communication application by touching the "Return" control 605.

[0113] In the above information processing method, by obtaining the abnormal description information specified by the Bayesian network and the target object in the current dialogue round, the abnormal category probability distribution corresponding to the abnormal description information can be determined based on the Bayesian network, and thus the first symptom category probability distribution corresponding to the abnormal category probability distribution can be determined based on the abnormal symptom transfer matrix. By obtaining the first symptom category probability distribution, the candidate symptom category that needs to be inquired in the next step can be accurately predicted based on the first symptom category probability distribution, thereby improving the accuracy of the candidate symptom category. By determining the candidate symptom category that needs to be inquired in the next step, the supplementary symptom category selected by the target object from the candidate symptom category can be obtained. In this way, the abnormal description information can be updated based on the supplementary symptom category, and the step of determining the abnormal category probability distribution corresponding to the abnormal description information based on the joint probability distribution between the abnormal category and the symptom category is returned and continued until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommendation information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommendation information based on the continuously updated abnormal description information, the efficiency of generating the recommendation information is greatly improved compared to manual screening.

[0114] In addition, since the present application prioritizes determining the probability value of the target object corresponding to each abnormal category when it has the current abnormal description information, and then obtains the probability value of the target object corresponding to each abnormal category when it has the current abnormal description information through the probability value of the target object corresponding to each abnormal category, compared with the traditional prediction of symptom category through symptom category, the prediction logic of the present application based on the probability value of each abnormal category to the probability value of each symptom category is more explainable, so that the probability distribution of the first symptom category determined based on this prediction logic can be more accurate.

[0115] In one embodiment, if the current conversation turn is not the first conversation turn, then in response to a selection operation on a candidate symptom category in the symptom category display area, a selected supplementary symptom category is obtained, including: for each conversation turn except the first conversation turn, the candidate symptom category determined by the previous conversation turn is displayed through the symptom category display area of ​​the current conversation turn; in response to a selection operation on a candidate symptom category in the symptom category display area, a selected supplementary symptom category is obtained.

[0116] Specifically, for each conversation round except the first conversation round, the medical communication application can display the candidate symptom categories determined by the previous conversation round through the symptom category display area. For example, Figure 2 202 is the symptom category display area in the dialogue interface, which displays the candidate symptom categories determined in the previous dialogue round, so that the target object can click on the candidate symptom categories displayed in the symptom category display area according to the symptom categories it has to obtain supplementary symptom categories.

[0117] In this embodiment, the candidate symptom categories are displayed in the symptom category display area, allowing the target user to directly select the corresponding supplementary symptom category without having to manually enter the supplementary symptom category, thereby improving the efficiency of determining the supplementary symptom category. In addition, the intuitive display of candidate symptom categories can also enhance the user experience.

[0118] In one embodiment, based on a Bayesian network between abnormal categories and symptom categories, determining the abnormal category probability distribution corresponding to the abnormal description information includes: obtaining a preset Bayesian network between the abnormal category and the symptom category, and determining the joint probability distribution between the abnormal category and the symptom category based on the Bayesian network; extracting the target symptom category in the abnormal description information, and the target symptom state corresponding to the target symptom category; determining the corresponding abnormal category probability distribution based on the target symptom category and the target symptom state and based on the joint probability distribution between the abnormal category and the symptom category; wherein the abnormal category probability distribution includes, when the target object has the target symptom category and the target symptom state, the probability value of the target object corresponding to each abnormal category.

[0119] Specifically, a Bayesian network between abnormality categories and symptom categories can be preset in the intelligent medical consultation system. Since the parameters of the Bayesian network are the joint probability distribution between the parent node and the connected child nodes, the intelligent medical consultation system can extract each parameter in the Bayesian network to obtain the joint probability distribution between abnormality categories and symptom categories. For the sake of convenience, the joint probability distribution between a parent node and the connected child nodes is referred to as the initial probability distribution below. Therefore, in the case of multiple parent nodes, the joint probability distribution between abnormality categories and symptom categories can include multiple initial probability distributions. For example, refer to Figure 4 When the parent node is "rhinitis", "upper respiratory tract infection", etc., and the child node is "allergy", "cough", "sore throat", etc., the initial probability distribution can be P1("rhinitis", "allergy", "cough", "sore throat", "dyspnea"), P2("pneumonia", "cough", "sore throat", "lung rales", "fever"), or P3("upper respiratory tract infection", "sore throat", "dyspnea", "lung rales", "fever"), so the joint probability distribution between the abnormality category and the symptom category is {p1, p2, p3}.

[0120] Furthermore, the intelligent medical consultation system extracts the target symptom category included in the abnormality description information, and the target symptom state corresponding to the target symptom category, so as to obtain the probability value of the target object corresponding to each abnormality category when the target object has the target symptom category and the target symptom state, that is, the abnormality category probability distribution, based on the target symptom category and the target symptom state and based on the joint probability distribution between the abnormality category and the symptom category.

[0121] In one embodiment, the abnormal description information of a non-first dialogue round may be composed of two lists. The first list stores the symptom categories and symptom states input by the target object in the first dialogue round. For example, when the target object complains of "headache, which has lasted for a while", the first list may be {"headache": "yes"}. The second list stores the supplementary symptom categories selected by the target object, and the symptom states of the supplementary symptom categories. For example, in the second dialogue round, when the target object selects "headache on one side", the intelligent medical consultation system stores "headache on one side" and the corresponding default symptom state in the second list, and obtains {"headache on one side": "yes"}. Furthermore, when the target object selects "dizziness" in the second dialogue round, the intelligent medical consultation system updates the second list to {"headache on one side": "yes", "dizziness": "yes"}.

[0122] When the abnormal description information of the current round is obtained, the medical communication application determines the first list and the second list corresponding to the abnormal description information of the current round, and uses the symptom categories stored in the first list and the second list as the target symptom categories, and uses the symptom states stored in the first list and the second list as the target symptom states of the corresponding target symptom categories.

[0123] In one embodiment, when the abnormality description information is a textual content, the medical communication application can segment the text and determine whether each segmented word corresponds to a symptom category and a symptom state. If a segmented word corresponds to a symptom category, it is determined as the target symptom category. If a segmented word corresponds to a symptom state, it is used as the target symptom state of the preceding target symptom category.

[0124] In the above embodiment, since the Bayesian network has a joint probability distribution between abnormality categories and symptom categories, an accurate abnormality category rate distribution can be obtained based on the target symptom category and target symptom state in the abnormality description information and according to the joint probability distribution.

[0125] In one embodiment, the joint probability distribution between abnormal categories and symptom categories includes multiple initial probability distributions; each initial probability distribution includes a probability distribution between the abnormal category and the associated associated symptom category; according to the target symptom category and the target symptom state, and based on the joint probability distribution between the abnormal category and the symptom category, the corresponding abnormal category probability distribution is determined, including: for each initial probability distribution in the multiple initial probability distributions, based on the probability distribution between the current abnormal category and the associated associated symptom category in the current initial probability distribution, the probability value of the current abnormal category appearing under the condition that the target symptom category and the target symptom state appear; the probability value of each abnormal category appearing is comprehensively calculated to obtain the abnormal category probability distribution.

[0126] Specifically, the joint probability distribution between abnormality categories and symptom categories can include multiple initial probability distributions, where each initial probability distribution is a probability distribution between a corresponding abnormality category and an associated symptom category. For example, as described above, the initial probability distribution P1 corresponds to the abnormality category "rhinitis," and the associated symptom categories associated with "rhinitis" are "allergies," "cough," "sore throat," and "dyspnea." That is, the abnormality category can be a parent node in a Bayesian network, and the associated symptom categories can be child nodes connected to the parent node.

[0127] Furthermore, for each of the multiple initial probability distributions, the medical communication application determines the probability of the current abnormality category occurring, given the presence of the target symptom category and target symptom state, based on the probability distribution between the current abnormality category and the associated symptom category in the current initial probability distribution. Using a unique elimination algorithm within the Bayesian network, the medical communication application then combines the probability values ​​of each abnormality category to generate a probability distribution for the abnormality category.

[0128] In one embodiment, reference Figure 7 , Figure 7 A schematic diagram of the initial probability distribution in one embodiment is shown, where a is the associated symptom category 1, b is the associated symptom category 2, c is the current abnormal category, and a 1 The symptom status of the associated symptom category 1 is "yes", a 2 The symptom status of the associated symptom category 1 is "none", a 3 The symptom status of the associated symptom category 1 is "uncertain", b 1 The symptom status of associated symptom category 2 is "yes", b 2 The symptom status of the associated symptom category 2 is "none", b 3 The symptom status of associated symptom category 2 is "uncertain", c 1 The abnormal status of the current abnormal category is "yes", that is, there is the current abnormal category, c 2 The abnormal state of the current abnormal category is "none", that is, there is no current abnormal state. Among them, each row of the initial probability distribution represents the probability value of the corresponding elements appearing at the same time. For example, the first row of the initial probability distribution can represent a 1 、b 1 and c 1 The probability of simultaneous occurrence is 0.25.

[0129] When the current initial probability distribution is obtained and the target symptom category and target symptom state of the target object are determined, the medical communication application can determine the probability value of the target object appearing in the current abnormal category based on the current initial probability distribution. For example, when the current abnormal category is c and the target symptom state of the target symptom category is b 1 When , the probability value of the target object determined by the medical communication application to have the current abnormal category is the sum of the probability values ​​in the 3rd, 7th and 11th rows, that is, 0.08+0+0.09=0.17.

[0130] In the above embodiment, it is only necessary to input the target symptom category and the target symptom state into the Bayesian network, so that the abnormal category probability distribution can be quickly obtained based on the Bayesian network, thereby improving the efficiency of determining the abnormal category probability distribution.

[0131] In one embodiment, the steps of generating a Bayesian network between abnormal categories and symptom categories include: obtaining a first data set corresponding to each abnormal category; the first data set includes at least one first co-occurrence information; the first co-occurrence information reflects the co-occurring abnormal categories, abnormal states of abnormal categories, symptom categories and symptom states of symptom categories; determining the associated symptom categories associated with each abnormal category based on multiple first data sets; for each abnormal category in the multiple abnormal categories, determining the initial probability distribution between the current abnormal category and the associated associated symptom category based on the first co-occurrence information in the current first data set corresponding to the current abnormal category; combining the initial probability distribution corresponding to each abnormal category to obtain a joint probability distribution between the abnormal category and the symptom category, and based on the joint probability distribution, obtaining a Bayesian network between the abnormal category and the symptom category.

[0132] Specifically, before generating the probability distribution of abnormal categories based on the Bayesian network, it is also necessary to build a Bayesian network. The terminal can determine the parent node required to build the Bayesian network, that is, determine multiple abnormal categories, and obtain the first data set corresponding to each abnormal category. Among them, the first data set includes at least one first co-occurrence information; the first co-occurrence information reflects the co-occurring abnormal categories, the abnormal states of the abnormal categories, the symptom categories and the symptom states of the symptom categories. For example, the first co-occurrence information can be "the target object has rhinitis, nasal congestion and high temperature", "the target object has rhinitis and nasal congestion, but does not have high temperature", "the target object does not have rhinitis, but has nasal congestion and high temperature", etc.

[0133] Among them, the abnormal state refers to the current state of the abnormal category. For example, when the abnormal category is "rhinitis", the corresponding abnormal state can be "yes", "no" or "uncertain". When the abnormal state is "yes", it can be characterized by rhinitis, and when the abnormal state is "no", it can be characterized by the absence of rhinitis. Correspondingly, the symptom state refers to the current state of the symptom category. It is easy to understand that when a first data set corresponds to an abnormal category, the first co-occurrence information in the first data set only records the abnormal category and the symptom category that co-occurs with the abnormal category. For example, when the first data set 1 corresponds to "rhinitis", the first co-occurrence information in the first data set 1 records the symptom category that co-occurs with "rhinitis".

[0134] Since each first data set corresponds to an abnormality category, and the first data set includes commonly occurring abnormality categories and symptom categories, the terminal can determine the associated symptom categories associated with each abnormality category through multiple first data sets. Among them, the associated symptom category associated with the abnormality category can be the symptom category that co-appears with the abnormality category. For example, when the first data set 1 corresponds to "rhinitis", and the first co-occurrence information in the first data set 1 records that the symptom categories that co-appear with "rhinitis" are "nasal congestion" and "high temperature", the terminal can determine that the associated symptom types associated with "rhinitis" are "nasal congestion" and "high temperature". In this way, the associated symptom categories associated with each abnormality category can be obtained.

[0135] Furthermore, for each of the multiple abnormality categories, the terminal determines an initial probability distribution between the current abnormality category and the associated symptom category based on the first co-occurrence information in the current first dataset corresponding to the current abnormality category, thereby using the initial probability distribution corresponding to each abnormality category as a parameter of the Bayesian network. The terminal uses the abnormality category as the parent node, the symptom category as the child node, and the initial probability distribution as the parameter to construct a Bayesian network between the abnormality category and the symptom category.

[0136] In one embodiment, the terminal may collect the first data set from the network, or obtain the first data set from a cooperative hospital, etc.

[0137] In one embodiment, the nodes in the Bayesian network can also use color to represent the current state of the node, for example, blue can represent the state of "no", orange can represent the state of "yes", white can represent the state of "uncertain", etc. It is easy to understand that the state can be an abnormal state or a symptomatic state.

[0138] In the above embodiment, by constructing a Bayesian network, the probability distribution of the abnormality category corresponding to the abnormality description information can be determined based on the Bayesian network.

[0139] In one embodiment, each abnormality category includes multiple abnormal states, and each associated symptom category includes multiple symptom states; based on the first co-occurrence information in the current first data set corresponding to the current abnormality category, the initial probability distribution between the current abnormality category and the associated associated symptom category is determined, including: determining multiple state combinations between the abnormal state of the current abnormality category and the symptom state of the corresponding associated symptom category; based on the first co-occurrence information of the current first data set, determining the number of occurrences of each state combination; based on the total number of first co-occurrence information in the current first data set and the number of occurrences of each state combination, obtaining the conditional probability value corresponding to each state combination; and combining the conditional probability values ​​corresponding to each state combination to obtain the initial probability distribution between the current abnormality category and the corresponding associated symptom category.

[0140] Specifically, since the abnormal category can have multiple abnormal states, the associated symptom category associated with the abnormal category can also have multiple symptom states. Therefore, when the current abnormal category and the associated symptom category associated with the current abnormal category are determined, the terminal can determine multiple state combinations between the abnormal state of the current abnormal category and the symptom state of the corresponding associated symptom category. For example, when the current abnormal category is "rhinitis", the associated symptom categories are "nasal congestion" and "high temperature", the abnormal state of the abnormal category can be "yes" or "no", and the symptom state of the associated symptom can be "yes" or "no", the multiple state combinations between the abnormal state of the current abnormal category and the symptom state of the corresponding associated symptom category can be (rhinitis = yes, nasal congestion = yes, high temperature = yes), (rhinitis = yes, nasal congestion = no, high temperature = yes), (rhinitis = yes, nasal congestion = no, high temperature = no), etc.

[0141] Furthermore, the terminal determines the number of occurrences of each state combination based on the first co-occurrence information of the current first data set. For example, based on the first co-occurrence information of the current first data set, the terminal determines that (rhinitis = yes, nasal congestion = yes, high temperature = yes) appears A times, determines that (rhinitis = yes, nasal congestion = no, high temperature = yes) appears B times, and so on. The terminal determines the total number of first co-occurrence information in the current first data set, and based on the total number and the number of occurrences of each state combination, obtains the conditional probability value corresponding to each of the state combinations. For example, the terminal may divide the total number by the number of occurrences of the corresponding state combination to obtain the conditional probability of the corresponding state combination. For example, when the total number is D, the conditional probability value corresponding to (rhinitis = yes, nasal congestion = yes, high temperature = yes) is A / D, and the conditional probability corresponding to (rhinitis = yes, nasal congestion = no, high temperature = yes) is B / D. In one embodiment, the terminal may add the number of occurrences corresponding to each state combination to obtain the total number of first co-occurrence information.

[0142] After obtaining the conditional probability value corresponding to each state combination, the terminal can integrate the conditional probability value corresponding to each state combination to obtain the initial probability distribution between the current abnormality category and the corresponding associated symptom category.

[0143] In this embodiment, by obtaining the first co-occurrence information in the current first data set, the conditional probability value corresponding to each state combination can be accurately obtained based on the first co-occurrence information, thereby obtaining an accurate initial probability distribution based on the accurate conditional probability value corresponding to each state combination.

[0144] In one embodiment, the steps of generating an abnormal symptom transfer matrix include: obtaining a second data set; the second data set includes at least one second co-occurrence information, and the second co-occurrence information reflects the co-occurring abnormality category and symptom category; for each abnormality category in multiple abnormality categories and each symptom category in multiple symptom categories, the total number of occurrences of the current abnormality category and the co-occurrence number of the current abnormality category and the current symptom category are determined based on the second co-occurrence information; based on the total number and the co-occurrence number, the conditional probability value between the current abnormality category and the current symptom category is determined; and the conditional probability values ​​between each abnormality category and the corresponding symptom category are synthesized to obtain the abnormal symptom transfer matrix.

[0145] Specifically, before generating the first symptom probability distribution through the abnormal symptom transfer matrix, an abnormal symptom transfer matrix can be constructed. The terminal can obtain the second data set required to construct the abnormal symptom transfer matrix, and the second data set includes at least one second co-occurrence information, and the second co-occurrence information records the abnormal categories and symptom categories that appear together. For example, the second co-occurrence information may be "rhinitis and nasal congestion appear together", "pneumonia and high temperature appear together", etc. Furthermore, for each abnormal category in the multiple abnormal categories and each symptom category in the multiple symptom categories, the terminal determines the total number of occurrences of the current abnormal category and the number of co-occurrences of the current abnormal category and the current symptom category based on the second co-occurrence information, and determines the conditional probability value between the current abnormal category and the current symptom category based on the total number and the number of co-occurrences. For example, when the abnormality category is "rhinitis" and the symptom categories are "cough" and "high temperature", the terminal can count the number of times "rhinitis" occurs, the number of times "rhinitis" and "cough" appear together, and the number of times "rhinitis" and "high temperature" appear together, and divide the number of times "rhinitis" and "cough" appear together by the number of times "rhinitis" appears to obtain the conditional probability value between "rhinitis" and "cough", and divide the number of times "rhinitis" and "high temperature" appear together by the number of times "rhinitis" appears to obtain the conditional probability value between "rhinitis" and "high temperature".

[0146] Furthermore, once the conditional probability values ​​between each abnormality category and the corresponding symptom category are obtained, the terminal integrates these conditional probability values ​​to obtain an abnormal symptom transfer matrix. In this matrix, different rows correspond to different abnormality categories, and different columns correspond to different symptom categories. Each element in the matrix represents the probability of the corresponding symptom category occurring under the condition that the corresponding abnormality category occurs.

[0147] In this embodiment, by generating an abnormal symptom transfer matrix, the probability distribution of the first symptom category can be subsequently determined based on the abnormal symptom transfer matrix.

[0148] In one embodiment, the above-mentioned information processing method also includes: extracting symptom categories and symptom states from the abnormal description information of the current dialogue round, and sorting the extracted symptom categories and symptom states to obtain structured data; generating a standard description text corresponding to the structured data; encoding the standard description text to obtain corresponding information features; outputting a second symptom category probability distribution corresponding to the information features through a perceptron model; the perceptron model is a network model obtained by training symptom labels corresponding to standard sample texts; determining candidate symptom categories based on the first symptom category probability distribution, including: comprehensively determining the candidate symptom categories by combining the first symptom category probability distribution and the second symptom category probability distribution.

[0149] Specifically, in order to improve the accuracy of the determined candidate symptom categories, the medical communication application can also generate a second symptom category probability distribution, and determine the final candidate symptom category by combining the first symptom category probability distribution and the second symptom category probability distribution. When the abnormal description information specified by the target object in the current dialogue round is obtained, the medical communication application can extract the symptom category and symptom status in the abnormal description information, and sort the extracted symptom category and symptom status according to the preset sorting rules to obtain structured data and generate a standard description text corresponding to the structured data. For example, when the abnormal description information is {headache: yes, diarrhea: no}, the medical communication application can sort "headache: yes" and "diarrhea: no" to obtain structured data with a fixed order "diarrhea + no + headache + yes", and output the corresponding standard description text "target object has no diarrhea but headache" for the structured data.

[0150] Furthermore, the medical communication application encodes the standard description text to obtain information features in the standard description text, and inputs the information features into a pre-trained perceptron model, which then outputs a second symptom category probability distribution. The perceptron model is a network model trained using symptom labels corresponding to the standard sample text. The perceptron model is a feedforward artificial neural network model commonly used for classification problems, which maps multiple input data sets to a label distribution, with each layer consisting of a linear layer and an activation layer.

[0151] When the first symptom category probability distribution and the second symptom category probability distribution are obtained, the medical communication application integrates the first symptom category probability distribution and the second symptom category probability distribution to obtain a candidate symptom category.

[0152] In this embodiment, candidate symptom categories are determined by combining the first symptom category probability distribution and the second symptom category probability distribution, making the determined candidate symptom categories more accurate. In addition, because text processing is very sensitive to word order, sorting the extracted symptom categories and symptom states to obtain a standard description text with a fixed symptom order can make the extracted information features more accurate, thereby making the second symptom category probability distribution determined based on the information features also more accurate.

[0153] In one embodiment, the first symptom category probability distribution and the second symptom category probability distribution are combined to determine candidate symptom categories, including: fusing each element in the first symptom category probability distribution with the element with the corresponding symptom category in the second symptom category probability distribution to obtain a target symptom category probability distribution; determining target elements in the target symptom category probability distribution whose probability values ​​meet high probability conditions, and using the symptom categories corresponding to each target element as candidate symptom categories.

[0154] Specifically, the medical communication application determines the symptom category corresponding to each element in the first symptom category probability distribution, and determines the symptom category corresponding to each element in the second symptom category probability distribution, and fuses each element in the first symptom category probability distribution with the element with the corresponding symptom category in the second symptom category probability distribution to obtain the target symptom category probability distribution. Furthermore, the medical communication application determines the target elements in the target symptom category probability distribution whose probability values ​​meet the high probability condition, and uses the symptom categories corresponding to each target element as candidate symptom categories. The high probability condition can be freely set according to needs. For example, symptom categories with probability values ​​higher than the probability threshold are used as candidate symptom categories.

[0155] In one embodiment, when the determined candidate symptom category is consistent with the symptom category in the abnormality description information, the medical communication application deletes the candidate symptom category and redetermines the candidate symptom category from the target symptom category probability distribution.

[0156] In one embodiment, the medical communication application may determine candidate symptom categories according to the formula S = argmax(P_s1 + P_s2). Here, P_s1 is the probability distribution of the first symptom category, and P_s2 is the probability distribution of the second symptom category. The medical communication application adds P_s2 to P_s1 to obtain the target symptom category probability distribution, and selects the symptom category with the maximum probability value in the target symptom category probability distribution as the candidate symptom category.

[0157] In the above embodiment, the candidate symptom categories are determined by combining the first symptom category probability distribution and the second symptom category probability distribution, so that the determined candidate symptom categories are more accurate.

[0158] In one embodiment, Figure 8 As shown, an information processing method is provided, which is applied to Figure 1 The following steps are used as an example to illustrate the terminal in the figure:

[0159] Step S802: Obtain the exception description information specified by the target object in the current dialogue round.

[0160] Step S804: Based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on the information features of the standard description text, a second symptom category probability distribution is determined.

[0161] Specifically, a medical communication application is running on the terminal. When generating recommendation information, the medical communication application can obtain the abnormality description information specified by the target party in the current conversation turn. The specific steps for the medical communication application to obtain the abnormality description information specified by the target party in the current conversation turn can be found in the specific embodiment corresponding to step S302.

[0162] Because the abnormality description information contains at least one symptom category and the corresponding symptom state for each symptom category, upon obtaining the abnormality description information, the medical communication application can sort the symptom categories and corresponding symptom states in the abnormality description information to obtain a standard description text with a fixed symptom order, and perform feature extraction on the standard description text to obtain information features. Furthermore, the medical communication application uses the extracted information features to determine the probability distribution of the second symptom category.

[0163] In one embodiment, the medical communication application may use BERT (Bidirectional Encoder Representations from Transformers) to extract information features from the standard description text and output the second symptom category probability distribution through a perceptron model.

[0164] In one embodiment, the medical communication application can determine the probability distribution of the second symptom category using the formula P_s2 = W*H + B. Here, H = BERT(X), where X is the standard description text, M is the information feature, Mean(M) represents the mean of the information feature, W is a learnable parameter in the model, and B is the bias term.

[0165] Step S806 , determining candidate symptom categories based on the second symptom category probability distribution, and entering the next dialogue round to obtain the supplementary symptom category selected by the target object from the candidate symptom categories.

[0166] Specifically, upon obtaining the second symptom category probability score, the medical communication application may determine candidate symptom categories based on the second symptom category probability distribution. The second symptom category probability distribution includes probability values ​​corresponding to each symptom category. For example, the medical communication application may select symptom categories that meet high probability conditions in the second symptom category probability distribution as candidate symptom categories. The second symptom category probability distribution in the current conversation turn refers to the probability values ​​of the target object for each symptom category when the target object corresponds to the abnormality description information in the current conversation turn.

[0167] Furthermore, the medical communication application displays the determined candidate symptom categories and enters the next conversation round, so that the target object can select a supplementary symptom category from the displayed candidate symptom categories.

[0168] Step S808, updates the abnormal description information based on the supplementary symptom category, returns to generate a standard description text with a fixed symptom order based on the abnormal description information, and continues to determine the second symptom probability distribution step based on the information features of the standard description text until the preset stop condition is reached.

[0169] Specifically, when the medical communication application displays candidate symptom categories, the next dialogue round can be entered. At this time, the target object can select a supplementary symptom category from the displayed candidate symptom categories, so that the medical communication application can update the acquired abnormality description information based on the selected supplementary symptom category to obtain the corresponding abnormality description information. In the next dialogue round, the medical communication application determines the second symptom probability distribution corresponding to the updated abnormality description information based on the BERT model and the perceptron model, and thus determines the candidate symptom category based on the second symptom category probability distribution. The medical communication application displays the determined candidate symptom category to enter the next dialogue round. And so on, until the preset stop condition is reached.

[0170] Step S810: Based on the finally updated abnormal description information, corresponding recommendation information is output to the target object.

[0171] Specifically, upon obtaining the final updated exception description, the medical communication application can generate and display corresponding recommendations based on the final updated exception description. Recommendations refer to information that pushes relevant content to the target audience. For example, recommendations could include recommended departments, recommended doctors, and recommended diets.

[0172] In the above information processing method, by obtaining the abnormal description information specified by the target object in the current dialogue round, a standard description text with a fixed symptom order corresponding to the abnormal description information can be generated, so that more accurate information features can be extracted from the standard description text with a fixed symptom order, and a more accurate second symptom category probability distribution can be generated based on the more accurate information features. By generating the second symptom category probability distribution, the candidate symptom category that needs to be inquired in the next step can be predicted based on the second symptom category probability distribution, so that the supplementary symptom category selected by the target object can be obtained through the candidate symptom category that needs to be inquired in the next step. In this way, the abnormal description information can be updated based on the supplementary symptom category, and the step of generating a standard description text with a fixed symptom order based on the abnormal description information and determining the second symptom probability distribution based on the information features of the standard description text is continued until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommended information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommended information based on the continuously updated abnormal description information, the efficiency of generating the recommended information is greatly improved compared to the traditional manual screening.

[0173] In addition, since the text processing process is very sensitive to word order, sorting the extracted symptom categories and symptom states to obtain a standard description text with a fixed symptom order can make the extracted information features more accurate, thereby making the probability distribution of the second symptom category determined based on the information features more accurate.

[0174] In one embodiment, based on the abnormal description information, a standard description text with a fixed symptom order is generated, including: extracting the symptom categories and symptom states in the abnormal description information of the current dialogue round, and sorting the extracted symptom categories and symptom states to obtain structured data; generating a standard description text with a fixed symptom order corresponding to the structured data.

[0175] Specifically, when the abnormal description information specified by the target object in the current conversation round is obtained, the medical communication application can extract the symptom category and symptom status in the abnormal description information, and sort the extracted symptom category and symptom status according to the preset sorting rules to obtain structured data and generate a standard description text corresponding to the structured data.

[0176] For example, if the abnormality description information is {headache: yes, diarrhea: no}, the medical communication application can normalize it into a standard text representation, resulting in normalized data consisting of "patient" + "yes" + "headache" + "no" + "diarrhea" + ".". Furthermore, because text processing models are sensitive to word order, before extracting information features, the medical communication application sorts the symptom categories in the normalized data to obtain structured data with a fixed symptom order. This allows the application to subsequently generate standard description text with a fixed symptom order based on this structured data.

[0177] In this embodiment, by generating a standard description text with a fixed symptom order, the extracted information features can be made more accurate, and thus the probability distribution of the second symptom category determined based on the information features can also be made more accurate.

[0178] In one embodiment, based on the information features of the standard description text, the second symptom category probability distribution is determined, including: encoding the standard description text to obtain the corresponding information features; outputting the second symptom category probability distribution corresponding to the information features through a perceptron model; the perceptron model is a network model obtained by training the symptom labels corresponding to the standard sample text.

[0179] Specifically, medical communication applications can encode standard description texts through the BERT model to obtain corresponding information features, and input the information features into a pre-trained perceptron model to output the second symptom category probability distribution through the perceptron model.

[0180] In one embodiment, before outputting the second symptom category probability distribution based on the perceptron model, the terminal may train the perceptron model. The terminal obtains standard sample text and symptom labels corresponding to the standard sample text, and continuously adjusts the model parameters of the perceptron model based on the standard sample text and the symptom labels corresponding to the standard sample text until a training stop condition is met.

[0181] In the above embodiments, since traditional technologies generally predict symptom category probability distribution based on one-hot encoding features, although the symptom category probability distribution can be obtained, the one-hot encoding features reflect 0 / 1 encoding features, which ignores the information contained in the entity text, resulting in a low accuracy of the predicted symptom category probability distribution. In response to the above defects, the embodiment of the present application converts the abnormal description information into a standard description text and extracts the information features in the standard description text. In this way, the second symptom category probability distribution generated by the information features containing text information can be made more accurate.

[0182] In one embodiment, Figure 9As shown, an information processing method is provided, which is applied to Figure 1 The following steps are used as an example to illustrate the terminal in the figure:

[0183] Step S902: Obtain the exception description information specified by the target object in the current dialogue round.

[0184] Step S904: determine an abnormality category probability distribution based on the abnormality description information, and determine a related first symptom category probability distribution based on the abnormality category probability distribution.

[0185] Specifically, upon obtaining the abnormality description information specified by the target object in the current conversation round, the medical communication application may determine the abnormality category probability distribution based on the abnormality description information, and determine the related first symptom category probability distribution based on the abnormality category probability distribution. Determining the abnormality category probability distribution based on the abnormality description information includes: determining the abnormality category probability distribution corresponding to the abnormality description information based on a Bayesian network between the abnormality category and the symptom category. Determining the related first symptom category probability distribution based on the abnormality category probability distribution includes: determining the first symptom category probability distribution based on the abnormal symptom transition matrix and the abnormality category probability distribution.

[0186] Step S906: Generate a standard description text with a fixed symptom order based on the abnormal description information, and determine the second symptom category probability distribution based on the information features of the standard description text.

[0187] Among them, based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on the information characteristics of the standard description text, the specific steps of determining the probability distribution of the second symptom category can refer to the embodiment of the above-mentioned step S804.

[0188] Step S908 : Determine candidate symptom categories based on the first symptom category probability distribution and the second symptom category probability distribution, and enter the next dialogue round to obtain the supplementary symptom category selected by the target object from the candidate symptom categories.

[0189] Specifically, in order to improve the accuracy of the determined candidate symptom categories, the medical communication application can also generate a second symptom category probability distribution, determine the final candidate symptom category by combining the first symptom category probability distribution and the second symptom category probability distribution, and enter the next conversation round to obtain the supplementary symptom category selected by the target object from the candidate symptom categories.

[0190] Step S910, update the abnormality description information based on the supplementary symptom category, return to the joint probability distribution between the abnormality category and the symptom category, and continue to execute the step of determining the abnormality category probability distribution corresponding to the abnormality description information until the preset stop condition is reached.

[0191] Step S912: Based on the finally updated abnormal description information, corresponding recommendation information is output to the target object.

[0192] Specifically, when the medical communication application displays candidate symptom categories, the next conversation round begins. The target user can then select a supplementary symptom category from the displayed candidate symptom categories. The medical communication application can then update the abnormality description information in the current conversation round based on the selected supplementary symptom category, obtaining the abnormality description information for the next conversation round. This process continues in this manner until a preset stopping condition is met.

[0193] Upon receiving the final updated exception description, the medical communication application can generate and display corresponding recommendations based on the final updated exception description. Recommendations refer to information that pushes relevant content to the target user. For example, recommendations could include recommended departments, recommended doctors, and recommended diets.

[0194] The above-mentioned information processing method can generate a first symptom category probability distribution and a second symptom category probability distribution corresponding to the abnormal description information by obtaining the abnormal description information specified by the target object in the current conversation round, so that the candidate symptom category that needs to be inquired about next can be accurately predicted based on the first symptom category probability distribution and the second symptom category probability distribution. By accurately predicting the candidate symptom category that needs to be inquired about next, the supplementary symptom category selected by the target object from the candidate symptom category can be obtained. In this way, the abnormal description information can be updated based on the supplementary symptom category until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommendation information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommendation information based on the continuously updated abnormal description information, the efficiency of generating recommendation information is greatly improved compared to traditional manual screening.

[0195] This application also provides an application scenario, which applies the above-mentioned information processing method. Specifically, the application of the information processing method in this application scenario is as follows:

[0196] When the target subject wishes to determine the department to register, the target subject can input the symptom category he or she has through the medical communication application, and select a supplementary symptom category from the candidate symptom categories displayed by the medical communication application, so that the medical communication application can determine the corresponding recommended department based on the symptom category complained by the target subject and the selected supplementary category. In this way, the target subject can go to the recommended department to register for treatment.

[0197] The present application also provides an application scenario, which applies the above-mentioned information processing method.

[0198] Specifically, the application of the information processing method in this application scenario is as follows:

[0199] When the target subject wishes to determine a recommended doctor, the target subject can input the symptom category he or she has through the medical communication application, and select a supplementary symptom category from the candidate symptom categories displayed by the medical communication application, so that the medical communication application can determine the corresponding recommended doctor based on the symptom category complained by the target subject and the selected supplementary category. In this way, the target subject can go to the recommended doctor for treatment.

[0200] The above application scenarios are merely illustrative. It will be understood that the application of the business-related data reporting methods provided in the various embodiments of the present application is not limited to the above scenarios.

[0201] In a specific embodiment, referring to Figure 10 , the information processing method comprises the following steps:

[0202] S1002, displaying a dialogue interface; the dialogue interface includes an information input box and a symptom category display area.

[0203] S1004, if the current dialogue round is the first dialogue round, then in response to the information input operation triggered by the information input box, the abnormal description information of the first dialogue round is obtained; if the current dialogue round is not the first dialogue round, then in response to the selection operation of the candidate symptom category in the symptom category display area, the selected supplementary symptom category is obtained, and the abnormal description information of the previous dialogue round is updated based on the supplementary symptom category to obtain the abnormal description information of the current dialogue round.

[0204] S1006 , obtaining a preset Bayesian network between the abnormality category and the symptom category, and determining a joint probability distribution between the abnormality category and the symptom category based on the Bayesian network.

[0205] S1008, extracting the target symptom category and the target symptom state corresponding to the target symptom category from the abnormality description information.

[0206] S1010, the joint probability distribution between abnormal categories and symptom categories includes multiple initial probability distributions; each initial probability distribution includes the probability distribution between the corresponding abnormal category and the associated associated symptom category; for each initial probability distribution in the multiple initial probability distributions, based on the probability distribution between the current abnormal category and the associated associated symptom category in the current initial probability distribution, the probability value of the current abnormal category appearing under the condition that the target symptom category and the target symptom state appear; the probability values ​​of each abnormal category appearing are combined to obtain the abnormal category probability distribution.

[0207] S1012, different matrix rows in the abnormal symptom transfer matrix correspond to different abnormal categories, and different matrix columns correspond to different symptom categories; each element in the abnormal symptom transfer matrix represents the probability value of the corresponding symptom category under the condition that the corresponding abnormal category occurs; the abnormal category probability distribution is fused with each matrix column in the abnormal symptom transfer matrix respectively to obtain the first symptom category probability distribution of the current dialogue round.

[0208] S1014, extract the symptom categories and symptom states from the abnormal description information of the current dialogue round, and sort the extracted symptom categories and symptom states to obtain structured data; generate a standard description text corresponding to the structured data; encode the standard description text to obtain corresponding information features.

[0209] S1016, outputting a second symptom category probability distribution corresponding to the information feature through a perceptron model; the perceptron model is a network model trained by symptom labels corresponding to standard sample texts.

[0210] S1018, fusing each element in the first symptom category probability distribution with the element with the corresponding symptom category in the second symptom category probability distribution to obtain a target symptom category probability distribution.

[0211] S1020, determine the target elements whose probability values ​​in the target symptom category probability distribution meet the high probability condition, and use the symptom categories corresponding to each target element as candidate symptom categories, and enter the next dialogue round to obtain the supplementary symptom categories selected by the target object from the candidate symptom categories.

[0212] S1022, update the abnormal description information based on the supplementary symptom category, return to the step of extracting the target symptom category in the abnormal description information, and the target symptom state corresponding to the target symptom category, and continue to execute until the preset stop condition is reached.

[0213] S1024: Based on the finally updated abnormal description information, output corresponding recommendation information to the target object.

[0214] The above-mentioned information processing method can generate a first symptom category probability distribution and a second symptom category probability distribution corresponding to the abnormal description information by obtaining the abnormal description information specified by the target object in the current conversation round, so that the candidate symptom category that needs to be inquired about next can be accurately predicted based on the first symptom category probability distribution and the second symptom category probability distribution. By accurately predicting the candidate symptom category that needs to be inquired about next, the supplementary symptom category selected by the target object from the candidate symptom category can be obtained. In this way, the abnormal description information can be updated based on the supplementary symptom category until the final updated abnormal description information is obtained. By obtaining the final updated abnormal description information, the corresponding recommendation information can be output to the target object based on the final updated abnormal description information. Since the present application can automatically output the corresponding recommendation information based on the continuously updated abnormal description information, the efficiency of generating recommendation information is greatly improved compared to traditional manual screening.

[0215] It should be understood that although Figure 3 、 Figures 8-10 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 3 、 Figures 8-10 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0216] In one embodiment, Figure 11 As shown, an information processing device 1100 is provided. The device can be a software module or a hardware module, or a combination of both to form a part of a computer device. The device specifically includes: an information acquisition module 1102, a supplementary symptom determination module 1104, and a recommendation information generation module 1106, wherein:

[0217] Information acquisition module 1102, used to obtain the abnormal description information specified by the target object in the current dialogue round;

[0218] Supplementary symptom determination module 1104 is configured to determine the probability distribution of the abnormality category corresponding to the abnormality description information based on a Bayesian network between the abnormality category and the symptom category, and determine the probability distribution of the first symptom category based on the abnormal symptom transfer matrix and the abnormality category probability distribution; determine the candidate symptom category based on the first symptom category probability distribution, enter the next dialogue round, and obtain the supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the joint probability distribution between the abnormality category and the symptom category, and continue to determine the probability distribution of the abnormality category corresponding to the abnormality description information until a preset stopping condition is reached;

[0219] The recommendation information generating module 1106 is configured to output corresponding recommendation information to the target object based on the finally updated abnormality description information.

[0220] In one embodiment, the information acquisition module 1102 is also used to display a dialogue interface; the dialogue interface includes an information input box and a symptom category display area; if the current dialogue turn is the first dialogue turn, then in response to the information input operation triggered by the information input box, the abnormal description information of the first dialogue turn is obtained; if the current dialogue turn is not the first dialogue turn, then in response to the selection operation of the candidate symptom category in the symptom category display area, the selected supplementary symptom category is obtained, and the abnormal description information of the previous dialogue turn is updated based on the supplementary symptom category to obtain the abnormal description information of the current dialogue turn.

[0221] In one embodiment, the supplementary symptom determination module 1104 also includes a probability distribution determination module 1141, which is used to obtain a Bayesian network between preset abnormality categories and symptom categories, and determine the joint probability distribution between the abnormality categories and the symptom categories based on the Bayesian network; extract the target symptom category in the abnormality description information, and the target symptom state corresponding to the target symptom category; determine the corresponding abnormality category probability distribution based on the target symptom category and the target symptom state, and based on the joint probability distribution between the abnormality category and the symptom category; wherein the abnormality category probability distribution includes, when the target object has the target symptom category and the target symptom state, the probability value of the target object corresponding to each abnormality category.

[0222] In one embodiment, the joint probability distribution between abnormal categories and symptom categories includes multiple initial probability distributions; each initial probability distribution includes the probability distribution between the corresponding abnormal category and the associated associated symptom category; the probability distribution determination module 1141 is also used to determine, for each initial probability distribution in the multiple initial probability distributions, the probability value of the current abnormal category appearing under the condition that the target symptom category and the target symptom state appear; the probability values ​​of each abnormal category appearing are combined to obtain the abnormal category probability distribution.

[0223] In one embodiment, different matrix rows in the abnormal symptom transfer matrix correspond to different abnormal categories, and different matrix columns correspond to different symptom categories; each element in the abnormal symptom transfer matrix represents the probability value of the corresponding symptom category under the condition that the corresponding abnormal category occurs; the supplementary symptom determination module 1104 also includes a first distribution determination module 1142, which is used to fuse the abnormal category probability distribution with each matrix column in the abnormal symptom transfer matrix to obtain the first symptom category probability distribution of the current dialogue round.

[0224] In one embodiment, the information processing device 1100 is also used to obtain a first data set corresponding to each abnormal category; the first data set includes at least one first co-occurrence information; the first co-occurrence information reflects the co-occurring abnormal categories, abnormal states of abnormal categories, symptom categories and symptom states of symptom categories; the associated symptom categories associated with each abnormal category are determined based on multiple first data sets; for each abnormal category in the multiple abnormal categories, the initial probability distribution between the current abnormal category and the associated associated symptom category is determined based on the first co-occurrence information in the current first data set corresponding to the current abnormal category; the initial probability distribution corresponding to each abnormal category is combined to obtain a joint probability distribution between the abnormal category and the symptom category, and based on the joint probability distribution, a Bayesian network between the abnormal category and the symptom category is obtained.

[0225] In one embodiment, each abnormality category includes multiple abnormal states, and each associated symptom category includes multiple symptom states; the information processing device 1100 is also used to determine multiple state combinations between the abnormal state of the current abnormality category and the symptom state of the corresponding associated symptom category; determine the number of occurrences of each state combination based on the first co-occurrence information of the current first data set; obtain the conditional probability value corresponding to each state combination based on the total number of first co-occurrence information in the current first data set and the number of occurrences of each state combination; and obtain the initial probability distribution between the current abnormality category and the corresponding associated symptom category by combining the conditional probability values ​​corresponding to each state combination.

[0226] In one embodiment, the information processing device 1100 is also used to obtain a second data set; the second data set includes at least one second co-occurrence information, and the second co-occurrence information reflects the co-occurring abnormality category and symptom category; for each abnormality category in multiple abnormality categories, and each symptom category in multiple symptom categories, the total number of times the current abnormality category occurs, and the number of co-occurrences of the current abnormality category and the current symptom category are determined based on the second co-occurrence information; based on the total number and the number of co-occurrences, the conditional probability value between the current abnormality category and the current symptom category is determined; and the conditional probability values ​​between each abnormality category and the corresponding symptom category are combined to obtain an abnormal symptom transfer matrix.

[0227] In one embodiment, the information processing device 1100 is also used to extract symptom categories and symptom states from the abnormal description information of the current dialogue round, and sort the extracted symptom categories and symptom states to obtain structured data; generate a standard description text corresponding to the structured data; encode the standard description text to obtain corresponding information features; output a second symptom category probability distribution corresponding to the information features through a perceptron model; the perceptron model is a network model obtained by training symptom labels corresponding to standard sample texts; and determine candidate symptom categories by combining the first symptom category probability distribution and the second symptom category probability distribution.

[0228] In one embodiment, the information processing device 1100 is also used to fuse each element in the first symptom category probability distribution with the element with the corresponding symptom category in the second symptom category probability distribution to obtain a target symptom category probability distribution; determine the target element in the target symptom category probability distribution whose probability value meets the high probability condition, and use the symptom category corresponding to each target element as a candidate symptom category.

[0229] In one embodiment, Figure 12 As shown, an information processing device 1200 is provided. The device can be a software module or a hardware module, or a combination of the two to form a part of a computer device. The device specifically includes: an abnormality determination module 1202, an information update module 1204 and an information output module 1206, wherein:

[0230] The exception determination module 1202 is used to obtain the exception description information specified by the target object in the current dialogue turn;

[0231] Information updating module 1204 is configured to generate a standard description text with a fixed symptom order based on the abnormality description information, determine a second symptom category probability distribution based on information features of the standard description text, determine candidate symptom categories based on the second symptom category probability distribution, enter the next dialogue round, and obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the steps of generating a standard description text with a fixed symptom order based on the abnormality description information, and determining a second symptom probability distribution based on information features of the standard description text, and continue executing until a preset stopping condition is reached;

[0232] The information output module 1206 is configured to output corresponding recommendation information to the target object based on the finally updated abnormality description information.

[0233] In one embodiment, the information update module 1204 is also used to extract the symptom categories and symptom states from the abnormal description information of the current dialogue round, and sort the extracted symptom categories and symptom states to obtain structured data; and generate a standard description text with a fixed symptom order corresponding to the structured data.

[0234] In one embodiment, the information update module 1204 is also used to encode the standard description text to obtain corresponding information features; output the second symptom category probability distribution corresponding to the information features through the perceptron model; the perceptron model is a network model obtained by training the symptom labels corresponding to the standard sample text.

[0235] In one embodiment, Figure 13 As shown, an information processing device 1300 is provided. The device can be a software module or a hardware module, or a combination of the two to form a part of a computer device. The device specifically includes: an information specifying module 1302, an information adjusting module 1304, and an output display module 1306, wherein:

[0236] The information specifying module 1302 is used to obtain the abnormal description information specified by the target object in the current dialogue round;

[0237] Information adjustment module 1304 is configured to determine an abnormality category probability distribution based on the abnormality description information, and determine a related first symptom category probability distribution based on the abnormality category probability distribution; generate a standard description text with a fixed symptom order based on the abnormality description information, and determine a second symptom category probability distribution based on information features of the standard description text; determine a candidate symptom category based on the first symptom category probability distribution and the second symptom category probability distribution, and enter the next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the joint probability distribution between the abnormality category and the symptom category, and continue to determine the abnormality category probability distribution corresponding to the abnormality description information until a preset stopping condition is reached;

[0238] The output display module 1306 is used to output corresponding recommendation information to the target object based on the finally updated abnormal description information.

[0239] For the specific definition of the information processing device, please refer to the definition of the information processing method above and will not be repeated here. Each module in the above-mentioned information processing device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0240] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 14 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an information processing method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0241] Those skilled in the art will understand that Figure 14The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0242] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0243] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0244] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of each of the above-described method embodiments.

[0245] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0246] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0247] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An information processing method, characterized in that: The method comprises: Get the exception description information specified by the target object in the current dialogue round; Determine the abnormality category probability distribution corresponding to the abnormality description information based on a Bayesian network between the abnormality category and the symptom category, and determine the first symptom category probability distribution according to the abnormal symptom transfer matrix and the abnormality category probability distribution; Determining candidate symptom categories based on the first symptom category probability distribution, and entering the next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; Updating the abnormality description information based on the supplementary symptom category, returning to the Bayesian network based on the abnormality category and the symptom category, and continuing to perform the step of determining the probability distribution of the abnormality category corresponding to the abnormality description information until a preset stopping condition is reached; Based on the finally updated abnormal description information, output corresponding recommendation information to the target object; Different matrix rows in the abnormal symptom transfer matrix correspond to different abnormal categories, and different matrix columns correspond to different symptom categories; each element in the abnormal symptom transfer matrix represents the probability value of the corresponding symptom category under the condition that the corresponding abnormal category occurs; Determining the first symptom category probability distribution according to the abnormal symptom transfer matrix and the abnormal category probability distribution includes: The abnormal category probability distribution is fused with each matrix column in the abnormal symptom transfer matrix to obtain the first symptom category probability distribution of the current dialogue round.

2. The method according to claim 1, characterized in that The obtaining of the exception description information specified by the target object in the current dialogue round includes: Display a dialogue interface; the dialogue interface includes an information input box and a symptom category display area; If the current dialogue round is the first dialogue round, in response to the information input operation triggered by the information input box, obtaining abnormal description information of the first dialogue round; If the current conversation round is not the first conversation round, then in response to the selection operation of the candidate symptom category in the symptom category display area, the selected supplementary symptom category is obtained, and the abnormal description information of the previous conversation round is updated based on the supplementary symptom category to obtain the abnormal description information of the current conversation round.

3. The method according to claim 1, characterized in that The determining of the probability distribution of abnormality categories corresponding to the abnormality description information based on a Bayesian network between abnormality categories and symptom categories includes: Obtaining a preset Bayesian network between abnormality categories and symptom categories, and determining a joint probability distribution between the abnormality categories and the symptom categories based on the Bayesian network; Extracting the target symptom category and the target symptom state corresponding to the target symptom category from the abnormal description information; According to the target symptom category and the target symptom state, and based on the joint probability distribution between the abnormal category and the symptom category, a corresponding abnormal category probability distribution is determined; wherein the abnormal category probability distribution includes, when the target object has the target symptom category and the target symptom state, the probability value of the target object corresponding to each abnormal category.

4. The method according to claim 3, characterized in that The joint probability distribution between the abnormality category and the symptom category includes a plurality of initial probability distributions; each of the initial probability distributions includes a probability distribution between the abnormality category and the associated associated symptom category; The determining of the corresponding abnormality category probability distribution based on the target symptom category and the target symptom state and on the joint probability distribution between the abnormality category and the symptom category includes: For each of the multiple initial probability distributions, determining a probability value of the current abnormality category occurring under the condition that the target symptom category and the target symptom state occur based on the probability distribution between the current abnormality category and the associated associated symptom category in the current initial probability distribution; The probability values ​​of each abnormal category are combined to obtain the abnormal category probability distribution.

5. The method according to any one of claims 1 to 4, characterized in that The steps of generating the Bayesian network between the abnormality category and the symptom category include: Obtaining a first data set corresponding to each abnormality category; the first data set includes at least one first co-occurrence information; the first co-occurrence information reflects the commonly occurring abnormality categories, abnormal states of the abnormality categories, symptom categories, and symptom states of the symptom categories; Determining, based on the plurality of first data sets, associated symptom categories associated with each abnormality category; For each abnormality category in the plurality of abnormality categories, determining an initial probability distribution between the current abnormality category and an associated associated symptom category based on first co-occurrence information in a current first data set corresponding to the current abnormality category; The initial probability distribution corresponding to each of the abnormal categories is synthesized to obtain a joint probability distribution between the abnormal category and the symptom category, and based on the joint probability distribution, a Bayesian network between the abnormal category and the symptom category is obtained.

6. The method according to claim 5, characterized in that in, Each of the abnormality categories includes multiple abnormal conditions, and each of the associated symptom categories includes multiple symptom conditions; The determining, based on the first co-occurrence information in the current first data set corresponding to the current abnormal category, an initial probability distribution between the current abnormal category and the associated associated symptom category includes: determining a plurality of state combinations between the abnormal state of the current abnormal category and the symptom state of the corresponding associated symptom category; Determining the number of occurrences of each state combination according to the first co-occurrence information of the current first data set; Obtaining a conditional probability value corresponding to each state combination according to the total amount of first co-occurrence information in the current first data set and the number of occurrences of each state combination; The conditional probability values ​​corresponding to each state combination are combined to obtain the initial probability distribution between the current abnormal category and the corresponding associated symptom category.

7. The method according to any one of claims 1 to 4, characterized in that The step of generating the abnormal symptom transfer matrix includes: Acquire a second data set; the second data set includes at least one second co-occurrence information, wherein the second co-occurrence information reflects a co-occurring abnormality category and a symptom category; For each abnormality category in the plurality of abnormality categories and each symptom category in the plurality of symptom categories, determining, based on the second co-occurrence information, a total number of occurrences of the current abnormality category and a co-occurrence number of occurrences of the current abnormality category and the current symptom category; Determining a conditional probability value between the current abnormality category and the current symptom category based on the total number of times and the co-occurrence number of times; The conditional probability values ​​between each abnormal category and the corresponding symptom category are combined to obtain the abnormal symptom transfer matrix.

8. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Extracting symptom categories and symptom states from the abnormal description information of the current conversation round, and sorting the extracted symptom categories and symptom states to obtain structured data; generating a standard description text corresponding to the structured data; Encoding the standard description text to obtain corresponding information features; Outputting a second symptom category probability distribution corresponding to the information feature through a perceptron model; the perceptron model is a network model trained by symptom labels corresponding to standard sample texts; Determining candidate symptom categories according to the first symptom category probability distribution includes: The first symptom category probability distribution and the second symptom category probability distribution are combined to determine a candidate symptom category.

9. The method according to claim 8, characterized in that The step of combining the first symptom category probability distribution and the second symptom category probability distribution to determine a candidate symptom category includes: fusing each element in the first symptom category probability distribution with the element with the corresponding symptom category in the second symptom category probability distribution to obtain a target symptom category probability distribution; Target elements whose probability values ​​in the target symptom category probability distribution meet a high probability condition are determined, and the symptom category corresponding to each target element is taken as a candidate symptom category.

10. An information processing method, characterized in that: The method comprises: Get the exception description information specified by the target object in the current dialogue round; Based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on information features of the standard description text, a second symptom category probability distribution is determined; determining a candidate symptom category based on the second symptom category probability distribution, and entering a next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; updating the abnormality description information based on the supplementary symptom category, returning the abnormality description information, generating a standard description text with a fixed symptom order, and continuing the step of determining a second symptom probability distribution based on information features of the standard description text until a preset stopping condition is reached; Based on the finally updated abnormal description information, output corresponding recommendation information to the target object; Generating a standard description text with a fixed symptom order based on the abnormal description information includes: Extracting symptom categories and symptom states from the abnormal description information of the current conversation round, and sorting the extracted symptom categories and symptom states to obtain structured data; generating a standard description text with a fixed symptom order corresponding to the structured data; Determining the second symptom category probability distribution based on the information features of the standard description text includes: Encoding the standard description text to obtain corresponding information features; The second symptom category probability distribution corresponding to the information feature is output through a perceptron model; the perceptron model is a network model trained by symptom labels corresponding to standard sample texts.

11. An information processing method, characterized in that: The method comprises: Obtaining abnormality description information specified by the target object in the current conversation round; determining an abnormality category probability distribution based on the abnormality description information, and determining a related first symptom category probability distribution based on the abnormality category probability distribution; Based on the abnormal description information, a standard description text with a fixed symptom order is generated, and based on information features of the standard description text, a second symptom category probability distribution is determined; Determining candidate symptom categories based on the first symptom category probability distribution and the second symptom category probability distribution, and entering the next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; The abnormality description information is updated based on the supplementary symptom category, and the step of returning the joint probability distribution between the abnormality category and the symptom category and determining the abnormality category probability distribution corresponding to the abnormality description information is continued until a preset stopping condition is reached; Based on the finally updated abnormal description information, output corresponding recommendation information to the target object; Generating a standard description text with a fixed symptom order based on the abnormal description information includes: Extracting symptom categories and symptom states from the abnormal description information of the current conversation round, and sorting the extracted symptom categories and symptom states to obtain structured data; generating a standard description text with a fixed symptom order corresponding to the structured data; Determining the second symptom category probability distribution based on the information features of the standard description text includes: Encoding the standard description text to obtain corresponding information features; The second symptom category probability distribution corresponding to the information feature is output through a perceptron model; the perceptron model is a network model trained by symptom labels corresponding to standard sample texts.

12. An information processing device, characterized in that: The device comprises: The information acquisition module is used to obtain the abnormal description information specified by the target object in the current dialogue round; a supplementary symptom determination module, configured to determine, based on a Bayesian network between abnormality categories and symptom categories, a probability distribution of abnormality categories corresponding to the abnormality description information, and determine, based on an abnormal symptom transfer matrix and the abnormality category probability distribution, a first symptom category probability distribution; determine, based on the first symptom category probability distribution, a candidate symptom category, and enter a next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return to the step of determining, based on the Bayesian network between abnormality categories and symptom categories, a probability distribution of abnormality categories corresponding to the abnormality description information, and continue until a preset stopping condition is reached; A recommendation information generation module, configured to output corresponding recommendation information to the target object based on the finally updated abnormal description information; Different matrix rows in the abnormal symptom transfer matrix correspond to different abnormal categories, and different matrix columns correspond to different symptom categories; each element in the abnormal symptom transfer matrix represents the probability value of the corresponding symptom category under the condition that the corresponding abnormal category occurs. The supplementary symptom determination module is also used to fuse the abnormal category probability distribution with each matrix column in the abnormal symptom transfer matrix to obtain the first symptom category probability distribution of the current dialogue round.

13. An information processing device, characterized in that: The device comprises: The exception determination module is used to obtain the exception description information specified by the target object in the current dialogue turn; an information updating module configured to generate a standard description text having a fixed symptom order based on the abnormal description information, and determine a second symptom category probability distribution based on information features of the standard description text; determine a candidate symptom category based on the second symptom category probability distribution, and enter a next dialogue round to obtain a supplementary symptom category selected by the target subject from the candidate symptom categories; update the abnormal description information based on the supplementary symptom category, and return the step of generating a standard description text having a fixed symptom order based on the abnormal description information and determining a second symptom probability distribution based on information features of the standard description text, and continue executing until a preset stopping condition is reached; An information output module is used to output corresponding recommendation information to the target object based on the abnormal description information finally updated; The information update module is also used to extract symptom categories and symptom states from the abnormal description information of the current dialogue round, and sort the extracted symptom categories and symptom states to obtain structured data; generate a standard description text with a fixed symptom order corresponding to the structured data; encode the standard description text to obtain corresponding information features; output the second symptom category probability distribution corresponding to the information features through a perceptron model; the perceptron model is a network model obtained by training the symptom labels corresponding to the standard sample text.

14. An information processing device, characterized in that: The device comprises: The information specification module is used to obtain the abnormal description information specified by the target object in the current dialogue round; An information adjustment module is configured to determine an abnormality category probability distribution based on the abnormality description information, and determine a related first symptom category probability distribution based on the abnormality category probability distribution; generate a standard description text with a fixed symptom order based on the abnormality description information, and determine a second symptom category probability distribution based on information features of the standard description text; determine a candidate symptom category based on the first symptom category probability distribution and the second symptom category probability distribution, and enter a next dialogue round to obtain a supplementary symptom category selected by the target object from the candidate symptom categories; update the abnormality description information based on the supplementary symptom category, return the joint probability distribution between the abnormality category and the symptom category, and continue to perform the step of determining the abnormality category probability distribution corresponding to the abnormality description information until a preset stopping condition is reached; An output display module is used to output corresponding recommendation information to the target object based on the abnormal description information finally updated; The information adjustment module is also used to extract symptom categories and symptom states from the abnormal description information of the current dialogue round, and sort the extracted symptom categories and symptom states to obtain structured data; generate a standard description text with a fixed symptom order corresponding to the structured data; encode the standard description text to obtain corresponding information features; output the second symptom category probability distribution corresponding to the information features through a perceptron model; the perceptron model is a network model obtained by training the symptom labels corresponding to the standard sample text.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.

16. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images