Intelligent question answering method and device, computer device and storage medium

Abstract

The application relates to the technical field of artificial intelligence, in particular to an intelligent question and answer method and device, computer equipment and a storage medium. The method comprises the following steps: in the case that an intelligent reply event is detected, target user data and a target question and answer analysis model are acquired; the target question and answer analysis model is deployed on a function calculation platform, and the target question and answer analysis model is obtained by updating an existing question and answer analysis model in the case that a business changes; and reply information is generated according to an analysis result of the target question and answer analysis model on the target user data. The application can improve the accuracy of intelligent question and answer.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an intelligent question-answering method, apparatus, computer device, and storage medium. Background Technology

[0002] Currently, customer service in various business systems mainly includes intelligent robot customer service and human online customer service. The "predictive response" in customer service systems refers to the ability to "predict" the user's scenario, trajectory, and operation. Before the user even asks a question, the customer service system can "guess" the user's question and provide a response.

[0003] Whether it's a human online customer service representative or an intelligent chatbot, the first step to achieving "pre-emptive response" is to quickly and completely obtain relevant information generated by the user's recent actions within the business system. When business operations change, the corresponding analysis strategies for user-related information will also change.

[0004] Currently, in scenarios with numerous and complex business system interactions, the updates to the intelligent question-answering analysis model are delayed, leading to a decrease in the accuracy of customer service's predictions of user questions. Therefore, improvements are urgently needed. Summary of the Invention

[0005] Therefore, it is necessary to provide an intelligent question-answering method, device, computer equipment, and storage medium that can improve the accuracy of question answering in response to the above-mentioned technical problems.

[0006] Firstly, this application provides an intelligent question-answering method, which includes:

[0007] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0008] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0009] In one embodiment, the target user data includes data in at least two dimensions, and the target question-answering analysis model includes at least two sub-models;

[0010] Accordingly, based on the analysis results of the target user data by the target question-answering analysis model, response information is generated, including:

[0011] Assign a sub-model to each dimension of data based on the business function of each sub-model.

[0012] Based on the analysis results of the data in the corresponding dimension of each sub-model, a response message is generated.

[0013] In one embodiment, updating an existing question-answering analysis model includes:

[0014] Based on the analyzable dimensional information generated by the business changes, and the analyzable dimensional information of the sample user data from the existing question-and-answer analysis model, determine the dimensional information to be adjusted.

[0015] Based on the sub-models corresponding to the dimensions to be adjusted, the existing question-answering analysis model is updated to obtain the target question-answering analysis model.

[0016] In one embodiment, the existing question-answering analysis model is updated based on the sub-model corresponding to the dimension information to be adjusted, to obtain the target question-answering analysis model, including:

[0017] If the existing question-answering analysis model has feature analysis functions in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0018] In one embodiment, if the sub-model corresponding to the dimension information to be adjusted is different from the sub-model corresponding to the existing question-answering analysis model only in the extracted data format, then the model update query information is output.

[0019] Based on the feedback information regarding the model update query, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, and update the existing question-answering analysis model.

[0020] In one embodiment, the method further includes:

[0021] Update the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

[0022] In one embodiment, the method further includes:

[0023] Determine the target response method based on the interaction patterns with the target users;

[0024] Based on the target response method, the reply information will be sent back to the target user.

[0025] Secondly, this application also provides an intelligent question-answering device, which includes:

[0026] The acquisition module is used to acquire target user data and target question-and-answer analysis model when a smart reply event is detected. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating the existing question-and-answer analysis model when business changes occur.

[0027] The response module is used to generate response information based on the analysis results of the target user data by the target question-and-answer analysis model.

[0028] Thirdly, this application also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0029] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0030] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0031] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0032] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0033] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0034] Fifthly, this application also provides a computer program product comprising a computer program that, when executed by a processor, performs the following steps:

[0035] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are obtained. The target question-and-answer analysis model is deployed on the Function Compute platform, and is obtained by updating the existing question-and-answer analysis model when business changes occur.

[0036] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0037] The aforementioned intelligent question-answering methods, devices, computer equipment, and storage media, through the Function Compute platform, publish question-answering analysis models. For developers, there is no need to worry about the construction and maintenance of the infrastructure layer, which can improve the deployment efficiency of question-answering analysis models. Therefore, after updating the question-answering analysis model based on business changes, the release time of the updated question-answering analysis model (i.e., the target question-answering analysis model) can be shortened. In turn, by timely publishing the updated question-answering analysis model, the target user data can be effectively analyzed in a timely manner using the updated question-answering analysis model, thereby improving the accuracy of response information generation. Attached Figure Description

[0038] Figure 1 This is a diagram illustrating the application environment of the intelligent question-answering method in one embodiment.

[0039] Figure 2 This is a flowchart illustrating an intelligent question-answering method in one embodiment;

[0040] Figure 3 This is a flowchart illustrating the process of matching each sub-model for each dimension in one embodiment;

[0041] Figure 4 This is a flowchart illustrating the process of providing feedback information based on the response method in one embodiment;

[0042] Figure 5 This is a flowchart illustrating the process of updating an existing question-answering analysis model in one embodiment;

[0043] Figure 6 This is a flowchart illustrating the process of updating an existing question-and-answer analysis model based on data format in one embodiment.

[0044] Figure 7 This is a flowchart illustrating the intelligent question-answering method in another embodiment;

[0045] Figure 8 This is a structural block diagram of an intelligent question-answering device in one embodiment;

[0046] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] The intelligent question-answering method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. For example, when server 104 detects a smart reply event, it obtains target user data and a target question-and-answer analysis model. The target question-and-answer analysis model is deployed on a function computing platform, and it is updated when business changes occur. Based on the analysis results of the target user data by the target question-and-answer analysis model, a reply message is generated. This allows the target user to receive the reply message through their terminal 102. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0049] Currently, customer service in various business systems mainly includes intelligent chatbot customer service and human online customer service. "Answering before the question is asked" in customer service systems refers to the ability to "predict" user scenarios, movements, and actions. The system can "guess" the user's question and provide a response before the user even asks it. Whether it's human online customer service or intelligent chatbot customer service, the first step to achieving "answering before the question is asked" is to quickly and completely obtain relevant information generated by the user's recent operations within the business system. For example, a bank's business includes diverse branches such as deposits, loans, investment consulting, wealth management, bancassurance, customer marketing, and e-commerce platforms, providing various services to users through these platforms. When business operations change, the intelligent question-answering analysis model that analyzes user-related information also changes accordingly. Currently, in scenarios with numerous and complex business system interactions, the update of intelligent question-answering analysis models is delayed, leading to reduced accuracy in predicting user questions. Therefore, improvements are urgently needed.

[0050] In one embodiment, such as Figure 2 As shown, an intelligent question-answering method is provided, which can be applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0051] S201, upon detecting an intelligent reply event, acquire target user data and the target question-and-answer analysis model.

[0052] In this embodiment, the so-called intelligent response event is an event triggered when the target user interacts with the customer service system integrated in server 104. The target user is any user interacting with the customer service system. In one possible implementation, the target user data can be the question information currently entered by the target user, or it can be historical question information entered by the target user within a historical time period, where the historical time period can be set according to the amount of data.

[0053] In another possible implementation, applications run on the business platform. Each business can correspond to multiple applications, and each application can also correspond to multiple businesses. The business platform provides various services to users via the network. After the applications are deployed, they process the business according to preset processing logic. Each business has its corresponding workflow, which involves various different work roles. For example, some business platforms include multiple roles such as customer service, operations and maintenance, decision center, testing, development, and downstream application maintenance. In this scenario, target user data can be the data generated when users operate on the corresponding applications on the business platform. This data is used to characterize the user's identity. For example, target user data can include basic identity data of the user in various dimensions and operation record data of the user in various dimensions.

[0054] Optionally, the target question-answering analysis model is obtained by updating an existing question-answering analysis model in the event of business changes. Furthermore, the target question-answering analysis model is deployed on the Function Compute platform.

[0055] For example, there is at least one target question-answering analysis model, and each target question-answering analysis model corresponds to the target user data. In one possible implementation, when the target user data is user question information, the corresponding target question-answering analysis model can be matched according to the content of the user question. In another possible implementation, when the target user data is data generated when the user performs operations on the corresponding application on the business platform, a candidate question-answering analysis model can be matched for each application, or the same candidate question-answering analysis model can be matched for all applications. In this case, the candidate question-answering analysis model corresponding to the application is only used as the target question-answering analysis model when the user generates corresponding data on the application.

[0056] Optionally, business changes are achieved through application changes. Business changes can refer to adjustments to the original application (such as changing application parameter settings or changing application execution logic), or the addition of new applications to the business platform.

[0057] Only after developers have developed the modified business logic and the corresponding question-and-answer analysis model can they use the modified model to analyze the data generated within the modified business logic. Therefore, each time a business logic changes, corresponding business change information is generated. This information instructs the Function Compute platform to deploy the modified business logic. For example, in one implementation, the business change information may include version information of the existing question-and-answer analysis model and the version information of the target question-and-answer analysis model; in another implementation, the business change information may represent the identifier and version number of the updated application, as well as the identifier and version number of the corresponding question-and-answer analysis model. In this scenario, when the Function Compute platform detects a business logic change, it locates the modified business logic and the corresponding question-and-answer analysis model based on the business change information, and deploys the modified question-and-answer analysis model to the Function Compute platform for users to use.

[0058] During the deployment of the modified question-and-answer analysis model to the Function Compute platform, server 104 automatically responds to the business change operation and publishes the modified question-and-answer analysis model on the Function Compute platform as a function instance. The Function Compute platform automates the deployment of the modified question-and-answer analysis model, shortening the model release time during the business change process. When users generate data on the updated application, the updated question-and-answer analysis model corresponding to the application is used as the target question-and-answer analysis model to perform timely and accurate analysis of the new data of the target user.

[0059] It's important to note that deploying the target question-answering analysis model by publishing function instances eliminates the need for developers to pre-deploy the model. Developers can focus solely on their core business logic: writing code for the modified question-answering analysis model, uploading it to the Function Compute platform, and configuring the corresponding function triggering rules. When a trigger event is received, the Function Compute platform, based on the developer's configured triggering rules, starts the container (i.e., execution environment) corresponding to the modified question-answering analysis model and executes it. Container creation and deletion are directly managed by the Function Compute platform and are not accessible to developers. Developers also do not need to manage various server performance metrics and resource utilization. For developers, there's no need to worry about the construction and maintenance of the infrastructure layer, improving service deployment efficiency and thus shortening the development time for business changes.

[0060] Furthermore, the Function Compute platform automatically scales up and down based on the peaks and troughs of user requests, utilizing or releasing underlying computing resources. Compared to the continuous online nature of question-answering analysis models in traditional technologies, the question-answering analysis models on the Function Compute platform are only loaded and executed when a request occurs, and are not continuously online, reducing the ineffective use of computing resources.

[0061] S202, Generate response information based on the analysis results of the target user data by the target question-and-answer analysis model.

[0062] Specifically, when the target user data is the question information input by the target user, the analysis result can be the answer template matched by the knowledge base corresponding to the target question-answering analysis model; when the target user data is the data generated when the user operates on the corresponding application on the business platform, the analysis result can be the evaluation of the user's data in various dimensions by the target question-answering analysis model, the generation of evaluation indicators for each dimension, and the generation of analysis results by combining the evaluation indicators and the corresponding text library.

[0063] Specifically, the response information can be a further processing of the analysis results. For example, the analysis results in text form can be converted into response information output in the form of a corresponding template, or they can be processed into voice information.

[0064] The aforementioned intelligent question-answering method, by publishing the question-answering analysis model through the Function Compute platform, eliminates the need for developers to concern themselves with the construction and maintenance of the infrastructure layer. This improves the deployment efficiency of the question-answering analysis model. Therefore, after updating the question-answering analysis model based on business changes, the release time of the updated question-answering analysis model (i.e., the target question-answering analysis model) can be shortened. Consequently, by promptly releasing the updated question-answering analysis model, the target user data can be effectively analyzed using the updated question-answering analysis model, thereby improving the accuracy of the generated response information.

[0065] When multiple applications on a business platform share a single target question-answering analysis model—for example, these applications could be e-commerce, car sales, or social networking—the data dimensions extracted by different business applications may differ. Therefore, to improve the accuracy of the target question-answering analysis model for data analysis, such as... Figure 3 As shown, the target user data includes data from at least two dimensions, and the target question-answering analysis model includes at least two sub-models. This embodiment provides an optional method for generating response information based on the analysis results of the target user data from the target question-answering analysis model, that is, it provides a way to refine S202. The specific implementation process may include:

[0066] S301 assigns a sub-model to each dimension of data based on the business function of each sub-model.

[0067] The user's data dimensions can include the user's basic information dimension and the user's operation information dimension. The business function of the sub-model refers to which dimension of data the sub-model is used to process.

[0068] For example, the user's basic information dimensions may include one or more of the following: demographic attribute data, social attribute data, account attribute data, and financial attribute data. Demographic attribute data may include: the user's age, gender, ethnicity, and social profile; social attribute data may include: the user's workplace and position; account attribute data may include: the user's account registration, account authentication, and login device data on the business platform; and financial attribute data may include: the user's asset data, loan data, wealth management data, investment data, and credit card data.

[0069] The user operation information dimension can include at least one of the following: access behavior data, social behavior data, account operation data, transaction behavior data, and risk tag data. Access behavior data can include actions such as changing name (ID), active scenario behavior, verification behavior, changing password, changing account binding, and deleting records. Social behavior data can include user chat behavior data, adding friends, and joining community behavior data. Transaction behavior data can include transaction amount, transaction frequency, transaction scenario, and transaction method data, with transaction methods including payment on behalf of others, receiving red envelopes, and instant payment. Risk tag data can include tags such as whether there are credit delinquencies, whether payment deductions have failed, and whether credit card repayments have failed.

[0070] Specifically, when assigning a sub-model to each dimension of data based on the business function of each sub-model, the assignment can be made according to a preset relationship table. This table stores each sub-model, its corresponding business function, and its corresponding data dimension. Therefore, once the business function of any sub-model is determined, the corresponding data dimension can be found in the preset relationship table.

[0071] It should be noted that the user information (including but not limited to user basic information and user operation information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0072] S302, based on the analysis results of the data of each sub-model for the corresponding dimension, generate response information.

[0073] The analysis results output by each sub-model can include the scoring indicators for the corresponding dimension. For example, the analysis results of the first sub-model represent a risk score of 10% for the target user, the analysis results of the second sub-model represent a loan score of S (amount greater than a preset upper limit), and the analysis results of the third sub-model represent a credit card score of I (usage frequency greater than a preset threshold). Correspondingly, the customer service system has a corresponding categorized text library configured for each dimension, and each categorized text library contains descriptive information corresponding to each level of scoring indicator. The customer service system generates response information based on the descriptive information for each dimension. In one possible implementation, the response information can be configured to be suitable for the target user's corresponding question-and-answer scenario, such as a human online customer service representative, an intelligent robot customer service representative, a voice question-and-answer scenario, a text question-and-answer scenario, etc.

[0074] In this embodiment, by setting up multiple sub-models to analyze data in various dimensions of the target user data, interference between data in different applications is reduced, and the accuracy of data analysis is improved.

[0075] Furthermore, without any business changes, the system only needs to receive pre-defined optimization instructions during use to improve the accuracy of the existing question-answering analysis model. Therefore, this intelligent question-answering method also includes: when the update conditions are met, the function computing platform updates the existing question-answering analysis model on the function computing platform based on the acquired training samples.

[0076] The update condition can be a timed event source. When the Function Compute platform receives the event source, it can train and update the existing question answering analysis model based on the training samples. The question answering analysis model can be updated on the Function Compute platform. The update of the target question answering analysis model can be to update all sub-models or to update some specified sub-models.

[0077] Correspondingly, in the event of a business change, the question-answering analysis model needs to be updated synchronously with the business change; in this case, in one embodiment, the intelligent question-answering method further includes: updating the version information of the target question-answering analysis model according to the version information of the existing question-answering analysis model.

[0078] The version information of the existing question-answering analysis model is the model identifier and version number of the question-answering analysis model before the update. For example, the version information of the existing question-answering analysis model is: model identifier is I, version number is 2.0.

[0079] Furthermore, based on the version information of the existing question-answering analysis model, updating the version information of the target question-answering analysis model includes: querying the updated question-answering analysis model in the version repository based on the model identifier 'I', and adjusting the model version of the updated question-answering model from version 2.0 to version 3.0; at this time, the version information of the updated question-answering analysis model is model identifier 'I' and version number 3.0; deploying this model on the Function Compute platform as a candidate question-answering analysis model, when the target user generates corresponding data on the business platform (application) corresponding to the candidate question-answering analysis model, using the candidate question-answering analysis model as the target question-answering analysis model, making the target question-answering analysis model the latest version of the question-answering analysis model, at this time, the version information of the target question-answering analysis model is: model identifier 'I' and version number 3.0.

[0080] Based on different question-and-answer needs, and in order to provide targeted services to different types of customers, further, such as Figure 4 As shown, this intelligent question-answering method also includes:

[0081] S401, Determine the target response method based on the interaction pattern with the target user.

[0082] The interaction modes between server 104 and users include a human customer service mode and a chatbot mode. After logging into the customer service system, users can choose their interaction mode. Specifically, the human customer service mode involves a human representative answering questions from the target customer, while the chatbot mode involves an automated chatbot answering questions from the target customer.

[0083] Furthermore, the response method for the human customer service mode is as follows: after confirming the response information generated by server 104, the customer service system sends the response information back to the human customer service representative, obtains the voice and / or text response output by the human representative, and then sends the voice and / or text response back to the target customer's terminal. The response method for the chatbot mode is as follows: after confirming the response information generated by server 104, the customer service system directly sends the response information to the target customer's terminal via voice or text.

[0084] S402, Based on the target response method, the reply information is sent back to the target user.

[0085] Optionally, after sending the reply information to the target user according to the target response method, obtain the target user's response information to the reply information; the target user's response information can be identified as a new intelligent reply event, and the target user data of the target user can be updated according to the target user's response information to predict the content that the user will ask next.

[0086] In this embodiment, by setting different interaction modes and response methods, diverse question-and-answer needs of users can be met.

[0087] When assigning a sub-model to each dimension of data based on the business function of each sub-model, the business function of the sub-model can be used not only to describe the corresponding dimension of the sub-model, but also to describe both the corresponding dimension and the application level. For example, a business platform may have applications A, B, and C. Application A contains users' loan information, investment information, and risk operation information; application B contains users' loan information, credit card information, and risk operation information; and application C contains users' credit card information and risk operation information. Because different applications extract data in different ways, different applications may need to configure different sub-models. In this case, the business functions of the sub-models describe the corresponding dimensions and applications. For example, the business functions of the above sub-models could be: {Application A, Loan Information}, {Application B, Loan Information}, {Application A, Investment Information}, {Application B, Credit Card Information}, {Application C, Credit Card Information}, {Application A, Risk Operation Information}, {Application B, Risk Operation Information}, and {Application C, Risk Operation Information}. When assigning sub-models to the data for each dimension, the application level must also be considered. That is, the sub-models corresponding to the above business functions could be: {Application A, Loan Information} corresponds to the first sub-model, {Application B, Loan Information} corresponds to the second sub-model, {Application A, Investment Information} corresponds to the third sub-model, {Application B, Credit Card Information} corresponds to the fourth sub-model, and {Application C, Credit Card Information} corresponds to the fifth sub-model; {Application A, Risk Operation Information} corresponds to the sixth sub-model, {Application B, Risk Operation Information} corresponds to the seventh sub-model, and {Application C, Risk Operation Information} corresponds to the eighth sub-model.

[0088] In this situation, such as Figure 5 As shown, this embodiment provides an optional method for updating an existing question-answering analysis model. The specific implementation process may include:

[0089] S501. Based on the dimension information to be analyzed generated by the business change, and the analyzable dimension information of the sample user data from the existing question-and-answer analysis model, determine the dimension information to be adjusted.

[0090] Among them, the changed business can be an application that has undergone business adjustments or a newly added application. Each business change has corresponding business change information.

[0091] For example, when the business change involves adjusting application A to include user loan information, investment information, risk operation information, and transaction behavior data, the existing question-and-answer analysis model on the Function Compute platform is updated accordingly. {Application A, Loan Information}, {Application A, Investment Information}, {Application A, Risk Operation Information}, and {Application A, Transaction Behavior Data} are designated as the analyzable dimensions generated by application A. The aforementioned {Application A, Loan Information}, {Application B, Loan Information}, {Application A, Investment Information}, {Application B, Credit Card Information}, {Application C, Credit Card Information}, {Application A, Risk Operation Information}, {Application B, Risk Operation Information}, and {Application C, Risk Operation Information} are identified as the analyzable dimensions of the existing question-and-answer analysis model for the sample user data. In this case, there is no sub-model for analyzing transaction behavior data or extracting transaction behavior data from application A among the analyzable dimensions; therefore, the dimension to be adjusted is determined to be {Application A, Transaction Behavior Data}.

[0092] For example, when the business is changed to a new application D, which includes users' financial information, and the existing question-and-answer analysis model on the function computing platform is updated, {Application D, Financial Information} is taken as the dimension information to be analyzed generated by Application D. The aforementioned {Application A, Loan Information}, {Application B, Loan Information}, {Application A, Financial Information}, {Application B, Credit Card Information}, {Application C, Credit Card Information}, {Application A, Risk Operation Information}, {Application B, Risk Operation Information}, and {Application C, Risk Operation Information} are taken as the analyzable dimension information of the existing question-and-answer analysis model for the sample user data. At this point, among the analyzable dimension information, there exists a sub-model (third sub-model) used for analyzing financial data, but no sub-model for extracting financial data from Application D. Therefore, the dimension information to be adjusted is determined to be {Application D, Financial Data}.

[0093] S502, based on the sub-model corresponding to the dimension information to be adjusted, update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0094] In the example above, when the dimension information to be adjusted is {Application A, Transaction Behavior Data}, the business function of the sub-model corresponding to this dimension information is: to extract transaction behavior data from Application A and analyze the transaction behavior data. When the dimension information to be adjusted is {Application D, Financial Management Data}, the business function of the sub-model corresponding to this dimension information is: to extract financial management data from Application D and analyze the financial management data.

[0095] After updating the sub-model corresponding to the changed business, developers can upload both the sub-model and the corresponding business function of the question-and-answer analysis model to the version repository as candidate sub-models. Then, the Function Compute platform queries the sub-models corresponding to the dimensions to be adjusted based on the business function among the candidate sub-models, and adds the sub-models corresponding to the dimensions to be adjusted to the existing question-and-answer analysis model to update the existing question-and-answer analysis model. When the target user generates corresponding data in the updated existing question-and-answer analysis model, the existing question-and-answer analysis model is determined as the target question-and-answer analysis model.

[0096] In the customer service system, taking the first and second sub-models mentioned above as examples, when analyzing loan information in application A using the first sub-model and loan information in application B using the second sub-model, the evaluation methods of the first and second sub-models may differ or be the same. In this embodiment, since the target customer's data dimensions are numerous, to facilitate a quick and unified evaluation of data in any dimension by the customer service system, the calculation methods of the first and second sub-models can be set to a unified method. In this case, the difference between the first and second sub-models lies only in the feature extraction method. To facilitate rapid updates of the corresponding sub-models in the target question-and-answer analysis model when business adjustments occur, each sub-model in this embodiment includes a feature extraction module and a feature analysis module.

[0097] This embodiment provides an optional method for updating an existing question-answering analysis model based on the sub-model corresponding to the dimension information to be adjusted, thereby obtaining a target question-answering analysis model. In other words, it provides a way to refine S502. The specific implementation process may include:

[0098] If the existing question-answering analysis model has feature analysis functions in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0099] As in the example above, taking the dimension information to be adjusted as {Application D, financial data} as an example, in the existing question-answering analysis model with the feature analysis function in the sub-model corresponding to the dimension information to be adjusted, there is a third sub-model for analyzing financial data. Therefore, at this time, it is only necessary to obtain the feature extraction module corresponding to Application D from each candidate sub-model in the version library, and bind the feature extraction module to the feature analysis function (feature analysis module) of the third sub-model. At this time, the feature analysis function (feature analysis module) of the third sub-model also corresponds to the feature extraction module corresponding to Application A and the feature extraction module of Application D, which simplifies the update process of the sub-model corresponding to the dimension information to be adjusted.

[0100] In this embodiment, when a business change occurs, for the case of extracting the same dimension data in different applications, the feature analysis function in the sub-model corresponding to the dimension information to be adjusted in the existing question-and-answer analysis model is reused, and only the feature extraction function of the sub-model corresponding to the changed business is updated, thereby realizing the function of quickly updating the existing question-and-answer analysis model.

[0101] To reuse the feature analysis functionality of the same sub-model, the input data for the features (feature analysis function) fed into the feature analysis module of that sub-model needs to have the same format. Therefore, such as Figure 6 As shown, this intelligent question-answering method also includes:

[0102] S601, if the sub-model corresponding to the dimension information to be adjusted is different from the corresponding sub-model in the existing question-answering analysis model only in the extracted data format, then output the model update query information.

[0103] For example, transaction behavior data may vary across different applications, with different transaction identifiers and representations of transaction objects, leading to differences in the format of the extracted features. In this case, an update query is generated to prompt developers to convert the data format extracted from the sub-models corresponding to the dimensions to be adjusted.

[0104] S602, based on the feedback information on the model update query information, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, and update the existing question answering analysis model.

[0105] Optionally, if the feedback information confirms the update, the Function Compute platform will release the sub-model after the delivery of the sub-model corresponding to the adjusted dimension information, which includes feature extraction and feature analysis functions.

[0106] If the feedback information indicates that no update is needed, then if the data format of the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is the same as the data format extracted by the corresponding sub-model in the existing question-answering analysis model, the Function Compute platform will only release the feature extraction function in the sub-model; if the data format of the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is different from the data format extracted by the corresponding sub-model in the existing question-answering analysis model, then the Function Compute platform will not release the feature extraction function in that sub-model, that is, it will not update the existing question-answering analysis model.

[0107] For example, based on the above embodiments, this embodiment provides an optional example of an intelligent question-answering method. For instance... Figure 7 As shown, the specific implementation process includes:

[0108] S701. Based on the dimension information to be analyzed generated by the business change, and the analyzable dimension information of the sample user data from the existing question-and-answer analysis model, determine the dimension information to be adjusted.

[0109] S702, If an existing question-answering analysis model has feature analysis functions in the sub-model corresponding to the dimension information to be adjusted, then execute S703.

[0110] S703, using the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, updates the existing question-answering analysis model to obtain the target question-answering analysis model;

[0111] S704, upon detecting a smart reply event, acquires target user data and the target question-and-answer analysis model;

[0112] In this process, a sub-model is assigned to each dimension of data based on the business function of each sub-model; the target question-answering analysis model is deployed on the function computing platform, and the target question-answering analysis model is obtained by updating the existing question-answering analysis model when business changes occur.

[0113] S705, Based on the analysis results of the target user data by the target question-and-answer analysis model, generate response information;

[0114] S706, Determine the target response method based on the interaction pattern with the target user;

[0115] S707, based on the target response method, the reply information is sent back to the target user.

[0116] The specific processes of S701-S707 described above can be found in the description of the above method embodiments. Their implementation principles and technical effects are similar, and will not be repeated here.

[0117] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0118] Based on the same inventive concept, this application also provides an intelligent question-answering device for implementing the intelligent question-answering method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more intelligent question-answering device embodiments provided below can be found in the limitations of the intelligent question-answering method described above, and will not be repeated here.

[0119] In one embodiment, such as Figure 8 As shown, an intelligent question-answering device 1 is provided, including: an acquisition module 11 and a response module 12, wherein:

[0120] The acquisition module 11 is used to acquire target user data and target question-and-answer analysis model when a smart reply event is detected; wherein, the target question-and-answer analysis model is deployed on the function computing platform, and the target question-and-answer analysis model is obtained by updating the existing question-and-answer analysis model when business changes occur;

[0121] The response module 12 is used to generate response information based on the analysis results of the target user data by the target question-and-answer analysis model.

[0122] In one embodiment, the target user data includes data from at least two dimensions, and the target question-and-answer analysis model includes at least two sub-models; the response module 12 includes:

[0123] The matching submodule is used to assign a submodel to each dimension of data based on the business function of each submodel;

[0124] The parsing submodule is used to generate response information based on the analysis results of the data in the corresponding dimension of each sub-model.

[0125] In one embodiment, the intelligent question-answering device further includes an update module, comprising:

[0126] The determination submodule is used to determine the dimension information to be adjusted based on the dimension information to be analyzed generated by the business changes and the analyzable dimension information of the sample user data from the existing question-and-answer analysis model.

[0127] The update submodule updates the existing question-answering analysis model based on the sub-model corresponding to the dimension information to be adjusted, thus obtaining the target question-answering analysis model.

[0128] In one embodiment, the update submodule is also used for:

[0129] If the existing question-answering analysis model has feature analysis functions in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0130] In one embodiment, the intelligent question-answering device further includes an inquiry module, which is used for:

[0131] If the sub-model corresponding to the dimension information to be adjusted is different from the corresponding sub-model in the existing question-answering analysis model only in the extracted data format, then the model update query information will be output.

[0132] Based on the feedback information regarding the model update query, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, and update the existing question-answering analysis model.

[0133] In one embodiment, the intelligent question-answering device further includes a matching module, which is used for:

[0134] Update the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

[0135] In one embodiment, the intelligent question-answering device further includes an interaction module, which is used for:

[0136] Determine the target response method based on the interaction patterns with the target users;

[0137] Based on the target response method, the reply information will be sent back to the target user.

[0138] Each module in the aforementioned intelligent question-answering device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0139] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores intelligent question-and-answer data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements an intelligent question-and-answer method.

[0140] Those skilled in the art will understand that Figure 9The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0142] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0143] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0144] In one embodiment, the target user data includes data in at least two dimensions, and the target question-answering analysis model includes at least two sub-models. When the processor executes the logic of the computer program to generate response information based on the analysis results of the target user data by the target question-answering analysis model, the following steps are specifically implemented: according to the business function of each sub-model, a sub-model is assigned to each dimension of data; and response information is generated based on the analysis results of each sub-model on the corresponding dimension of data.

[0145] In one embodiment, when the processor executes the logic of updating the existing question-answering analysis model using a computer program, the following steps are specifically implemented: determining the dimension information to be adjusted based on the dimension information to be analyzed generated by the change in business and the analyzable dimension information of the sample user data in the existing question-answering analysis model; updating the existing question-answering analysis model according to the sub-model corresponding to the dimension information to be adjusted, to obtain the target question-answering analysis model.

[0146] In one embodiment, when the processor executes a computer program to update an existing question-answering analysis model based on the sub-model corresponding to the dimension information to be adjusted, and obtains the target question-answering analysis model, the following steps are specifically implemented: If the existing question-answering analysis model has the feature analysis function in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model and obtain the target question-answering analysis model.

[0147] In one embodiment, when the processor executes the computer program, it also performs the following steps: if the sub-model corresponding to the dimension information to be adjusted is different from the sub-model corresponding to the existing question-answering analysis model only in the extracted data format, then output the model update query information.

[0148] Based on the feedback information regarding the model update query, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, and update the existing question-answering analysis model.

[0149] In one embodiment, when the processor executes the computer program, it further performs the following steps: updating the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

[0150] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the target response method based on the interaction pattern with the target user; and feeding back the response information to the target user based on the target response method.

[0151] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0152] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0153] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0154] In one embodiment, the target user data includes data in at least two dimensions, and the target question-answering analysis model includes at least two sub-models;

[0155] In one embodiment, when the logic of generating response information based on the analysis results of the target user data by the target question-answering analysis model is executed by the processor, the following steps are specifically implemented: according to the business function of each sub-model, a sub-model is assigned to the data of each dimension; and response information is generated based on the analysis results of the corresponding dimension of the data by each sub-model.

[0156] In one embodiment, when the logic for updating an existing question-and-answer analysis model is executed by a processor, the following steps are specifically implemented: determining the dimension information to be adjusted based on the dimension information to be analyzed generated by the change in business and the analyzable dimension information of the sample user data in the existing question-and-answer analysis model; updating the existing question-and-answer analysis model according to the sub-model corresponding to the dimension information to be adjusted, to obtain the target question-and-answer analysis model.

[0157] In one embodiment, when the logic of updating the existing question-answering analysis model based on the sub-model corresponding to the dimension information to be adjusted, and obtaining the target question-answering analysis model, is executed by the processor, the following steps are specifically implemented: If the existing question-answering analysis model has the feature analysis function in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0158] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: if the sub-model corresponding to the dimension information to be adjusted is different from the sub-model in the existing question-answering analysis model only in the extracted data format, then output model update query information; based on the feedback information on the model update query information, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted to update the existing question-answering analysis model.

[0159] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: updating the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

[0160] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target response method based on the interaction pattern with the target user; and feeding back the response information to the target user based on the target response method.

[0161] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0162] Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. The target question-and-answer analysis model is deployed on the Function Compute platform and is obtained by updating an existing question-and-answer analysis model in the event of business changes.

[0163] Based on the analysis results of the target user data by the target question-answering analysis model, response information is generated.

[0164] In one embodiment, the target user data includes data in at least two dimensions, and the target question-answering analysis model includes at least two sub-models;

[0165] In one embodiment, when the logic of generating response information based on the analysis results of the target user data by the target question-answering analysis model is executed by the processor, the following steps are specifically implemented: according to the business function of each sub-model, a sub-model is assigned to the data of each dimension; and response information is generated based on the analysis results of the corresponding dimension of the data by each sub-model.

[0166] In one embodiment, when the logic for updating an existing question-and-answer analysis model is executed by a processor, the following steps are specifically implemented: determining the dimension information to be adjusted based on the dimension information to be analyzed generated by the change in business and the analyzable dimension information of the sample user data in the existing question-and-answer analysis model; updating the existing question-and-answer analysis model according to the sub-model corresponding to the dimension information to be adjusted, to obtain the target question-and-answer analysis model.

[0167] In one embodiment, when the logic of updating the existing question-answering analysis model based on the sub-model corresponding to the dimension information to be adjusted, and obtaining the target question-answering analysis model, is executed by the processor, the following steps are specifically implemented: If the existing question-answering analysis model has the feature analysis function in the sub-model corresponding to the dimension information to be adjusted, then the feature extraction function in the sub-model corresponding to the dimension information to be adjusted is used to update the existing question-answering analysis model to obtain the target question-answering analysis model.

[0168] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: if the sub-model corresponding to the dimension information to be adjusted is different from the sub-model in the existing question-answering analysis model only in the extracted data format, then output model update query information; based on the feedback information on the model update query information, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted to update the existing question-answering analysis model.

[0169] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: updating the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

[0170] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the target response method based on the interaction pattern with the target user; and feeding back the response information to the target user based on the target response method.

[0171] It should be noted that the user information (including but not limited to user basic information and user operation information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0174] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An intelligent question-answering method, characterized in that, The method includes: Upon detecting a smart reply event, target user data and a target question-and-answer analysis model are acquired. This target question-and-answer analysis model is deployed on a function computing platform. The model is developed in response to business changes by determining the dimensions to be adjusted based on the changes in the business and the analyzable dimensions of the sample user data from existing question-and-answer analysis models. If the existing question-and-answer analysis model possesses feature analysis functionality in the sub-model corresponding to the dimension to be adjusted, then the feature extraction functionality in that sub-model is used to update the existing model. The function computing platform automatically scales elastically based on the peaks and troughs of user requests. The target user data includes data from at least two dimensions, and the target question-answering analysis model includes at least two sub-models; based on the business function of each sub-model, one sub-model is assigned to each dimension of data. Based on the analysis results of the data in the corresponding dimension of each sub-model, a response message is generated; The method further includes: If the sub-model corresponding to the dimension information to be adjusted is different from the sub-model corresponding to the existing question-answering analysis model only in the extracted data format, then output the model update query information. Based on the feedback information regarding the model update query information, determine whether to use the feature extraction function in the sub-model corresponding to the dimension information to be adjusted, and update the existing question-answering analysis model.

2. The method according to claim 1, characterized in that, The method further includes: Update the version information of the target question-answering analysis model based on the version information of the existing question-answering analysis model.

3. The method according to claim 1, characterized in that, The method further includes: Determine the target response method based on the interaction patterns with the target users; The response information will be sent back to the target user according to the target response method.

4. An intelligent question-and-answer device, characterized in that, The device is used to perform the intelligent question-answering method according to any one of claims 1-3, including: The acquisition module is used to acquire target user data and a target question-and-answer analysis model when a smart reply event is detected; wherein, the target question-and-answer analysis model is deployed on the function computing platform, and the target question-and-answer analysis model is obtained by updating an existing question-and-answer analysis model when business changes occur; The response module is used to generate response information based on the analysis results of the target user data by the target question-and-answer analysis model.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

Images