Service processing method and device, computer readable storage medium and electronic equipment

Abstract

The invention relates to the technical field of artificial intelligence, and provides a service processing method, a service processing device, a computer readable storage medium and an electronic device.The service processing method is applied to a trained service processing model, and the trained service processing model comprises a center model and a plurality of auxiliary models. The plurality of auxiliary models are associated with a plurality of service plates, different auxiliary models are associated with different service plates, and the method comprises the following steps: after receiving service demand information of a user, the center model determines a user intention according to the service demand information; the central model determines a to-be-called target auxiliary model according to the user intention, so that the target auxiliary model feeds back business demand data associated with the user intention; and the central model processes the business demand data to obtain a service processing result. The service processing efficiency can be improved.

Description

Background Art

[0002] As the core system of the BSS domain (Business Support System), the billing system involves many customer services provided by telecom operators. It must not only accurately calculate user fees, but also recover funds in a timely manner, minimize arrears, and improve fund utilization.

[0003] With the continuous development of telecommunications services, the services are becoming more and more complex, and the billing system is also becoming more and more complex, resulting in a large number of inquiries and complaints from users. Simple questions will be answered by account managers or customer service who have direct contact with customers, while some complex questions that require analysis are transferred to second-line billing operation and maintenance personnel for manual analysis and processing, which consumes a lot of labor costs and relies heavily on manual experience, resulting in low efficiency.

[0004] In view of this, there is an urgent need in the art to develop a new service processing method and device.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure. Summary of the invention

[0006] The purpose of the present disclosure is to provide a service processing method, a service processing device, a computer-readable storage medium and an electronic device, thereby overcoming the technical problem of low efficiency caused by the limitations of related technologies at least to a certain extent.

[0007] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0008] According to a first aspect of the present disclosure, a service processing method is provided, which is applied to a trained service processing model, wherein the trained service processing model includes a central model and multiple auxiliary models, wherein the multiple auxiliary models are associated with multiple business segments, and different auxiliary models are associated with different business segments, and the method includes:

[0009] After receiving the service demand information of the user, the central model determines the user intention according to the service demand information;

[0010] The central model determines the target auxiliary model to be called according to the user intention, so that the target auxiliary model feeds back the business demand data associated with the user intention;

[0011] The central model processes the business demand data to obtain a service processing result.

[0012] In an exemplary embodiment of the present disclosure, the central model determines the target auxiliary model to be called according to the user intention, including:

[0013] The central model splits the user intent to obtain at least one refined intent after the split;

[0014] The target auxiliary model to be called is determined according to the target business segment associated with each of the refined intentions.

[0015] In an exemplary embodiment of the present disclosure, each of the business segments includes a plurality of sub-business segments, and each of the sub-business segments is associated with a sub-model;

[0016] When there are multiple target auxiliary models to be called, calling the target auxiliary model and obtaining the business demand data fed back by the target auxiliary model includes:

[0017] The central model decides the order of calling the plurality of target auxiliary models;

[0018] Calling each of the target auxiliary models in the calling order, and calling at least one target sub-model associated with a target sub-business segment according to each of the detailed intents through each of the target auxiliary models to obtain business demand data for each of the detailed intents;

[0019] The business demand data is fed back to the central model through each of the target auxiliary models.

[0020] In an exemplary embodiment of the present disclosure, the processing of the business demand data to obtain a service processing result includes:

[0021] Integrate the plurality of business demand data to obtain a preliminary processing result;

[0022] Extracting key information from the preliminary processing results to obtain intermediate processing results;

[0023] The intermediate processing result is reorganized according to the pre-learned template speech to obtain the service processing result.

[0024] In an exemplary embodiment of the present disclosure, after obtaining the service processing result, the method further includes:

[0025] The central model reports the service processing result to the operation and maintenance personnel, so that the operation and maintenance personnel feed back the service processing result to the user.

[0026] In an exemplary embodiment of the present disclosure, the trained service processing model is obtained by:

[0027] Obtain a trained central model and multiple trained auxiliary models;

[0028] The trained central model and multiple trained auxiliary models are jointly trained to obtain the trained service processing model.

[0029] In an exemplary embodiment of the present disclosure, the trained central model is obtained by training in the following manner:

[0030] Collecting a first training sample; the first training sample includes a service demand information sample of a sample user, a service processing result sample for the service demand information sample, and a business data sample of the sample user for the multiple business sectors;

[0031] Inputting the first training sample into the central model to be trained, and outputting a first action strategy for the service demand information sample through the central model to be trained; the first action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the service processing result sample for the service demand information sample;

[0032] Determine a first reward value for the first action strategy according to a preconfigured first reward function;

[0033] When the first reward value reaches a first training termination condition, the training of the central model to be trained is terminated to obtain the trained central model.

[0034] In an exemplary embodiment of the present disclosure, each trained auxiliary model is trained in the following manner:

[0035] Collecting a second training sample; the second training sample includes a refined intention sample of a sample user, a business demand data sample for the refined intention sample, and a business data sample of the sample user for each of the business sectors;

[0036] The second training sample is input into each auxiliary model to be trained, and each auxiliary model to be trained outputs a second action strategy for the refined intent sample; the second action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the business demand data sample for the refined intent sample;

[0037] Determine a second reward value for the second action strategy according to a preconfigured second reward function;

[0038] When the second reward value reaches the second training termination condition, the training of each auxiliary model to be trained is terminated to obtain each trained auxiliary model.

[0039] According to a second aspect of the present disclosure, a service processing device is provided, wherein a trained service processing model is configured on the device, wherein the trained service processing model includes a central model and multiple auxiliary models, wherein the multiple auxiliary models are associated with multiple business segments, and different auxiliary models are associated with different business segments, wherein the device includes:

[0040] A user intention determination module is used to determine the user intention according to the service demand information through the central model after receiving the user's service demand information;

[0041] An auxiliary model calling module, used to determine a target auxiliary model to be called according to the user intention through the central model, so that the target auxiliary model feeds back business demand data associated with the user intention;

[0042] The processing result determination module is used to process the business demand data through the central model to obtain a service processing result.

[0043] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the service processing method described in the first aspect is implemented.

[0044] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the service processing method described in the first aspect above by executing the executable instructions.

[0045] It can be seen from the above technical solutions that the service processing method, service processing device, computer-readable storage medium and electronic device in the exemplary embodiments of the present disclosure have at least the following advantages and positive effects:

[0046] In the technical solutions provided in some embodiments of the present disclosure, after receiving the user's service demand information, the central model determines the user's intention based on the service demand information, and then determines the target auxiliary model to be called based on the user's intention, calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model, and finally processes the business demand data to obtain the service processing result. On the one hand, by analyzing the user's intention, the central model can better understand the user's needs, quickly locate the corresponding auxiliary model, and provide personalized service processing results by calling the relevant auxiliary model; on the other hand, through the collaboration between the central model and the auxiliary model, the service processing results for the user's service demand information are quickly output, thereby solving the problem of low manual efficiency of billing and operation personnel in related technologies when handling customer inquiries and complaints, and improving the service processing efficiency of various customer issues, so that users can get quick and accurate responses, and improve satisfaction and service quality.

[0047] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0049] Figure 1 A schematic diagram showing a flow chart of a service processing method in an embodiment of the present disclosure;

[0050] Figure 2 A flowchart showing how to obtain a trained service processing model in an embodiment of the present disclosure;

[0051] Figure 3 A schematic diagram showing a flow chart of how to obtain the trained central model in the embodiment of the present disclosure;

[0052] Figure 4 A flowchart showing how to obtain each trained auxiliary model in an embodiment of the present disclosure;

[0053] Figure 5 A schematic diagram showing the structure of a trained service processing model in an embodiment of the present disclosure is shown;

[0054] Figure 6 A schematic diagram showing the structure of a control unit in an embodiment of the present disclosure is shown;

[0055] Figure 7A flow chart showing how the central model calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model in the embodiment of the present disclosure;

[0056] Figure 8 A flowchart showing how to obtain a service processing result in an embodiment of the present disclosure;

[0057] Fig. 9 A schematic diagram showing a flow chart of a service processing method in an embodiment of the present disclosure for a specific application scenario;

[0058] Fig.10 A schematic diagram showing the structure of a service processing device in an exemplary embodiment of the present disclosure;

[0059] Fig.11 A schematic structural diagram of an electronic device in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0060] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0061] The terms "a", "an", "the" and "said" are used in this specification to indicate the presence of one or more elements / components / etc.; the terms "including" and "having" are used to express an open-ended inclusion and mean that additional elements / components / etc. may exist in addition to the listed elements / components / etc.; the terms "first" and "second" etc. are used only as labels and are not intended to limit the quantity of their objects.

[0062] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and their repeated descriptions will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0063] In response to the low manual efficiency of billing operation and maintenance personnel in handling customer inquiries and complaints, this paper creatively proposes a billing query system based on multi-level agents. This system fully utilizes the excellent generalization ability of large models in general task processing, as well as the high efficiency of automatic planning, decision-making and execution of complex tasks. By innovatively combining large models and AI agent technology, the traditional manual query and troubleshooting process is intelligently reconstructed, significantly improving the efficiency and accuracy of billing queries.

[0064] In the embodiments of the present disclosure, a service processing method is firstly provided, which overcomes the defect of low efficiency in the related art at least to a certain extent.

[0065] Figure 1 A flow chart of a service processing method in an embodiment of the present disclosure is shown. The execution subject of the service processing method may be a server that performs service processing.

[0066] refer to Figure 1 According to an embodiment of the present disclosure, a service processing method includes the following steps:

[0067] Step S110, after receiving the user's service demand information, the central model determines the user's intention according to the service demand information;

[0068] Step S120, the central model determines the target auxiliary model to be called according to the user intention, calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model;

[0069] Step S130: the central model processes the business demand data to obtain a service processing result.

[0070] exist Figure 1 In the technical solution provided by the illustrated embodiment, after receiving the user's service demand information, the central model determines the user's intention based on the service demand information, and then determines the target auxiliary model to be called based on the user's intention, calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model, and finally processes the business demand data to obtain the service processing result. On the one hand, by analyzing the user's intention, the central model can better understand the user's needs, quickly locate the corresponding auxiliary model, and provide personalized service processing results by calling the relevant auxiliary model; on the other hand, through the collaboration between the central model and the auxiliary model, the service processing results for the user's service demand information are quickly output, thereby solving the problem of low manual efficiency of billing and operation personnel in related technologies when handling customer inquiries and complaints, and improving the service processing efficiency of various customer issues, so that users can get quick and accurate responses, improving satisfaction and service quality.

[0071] The following Figure 1 The specific implementation process of each step is described in detail:

[0072] Before step S110, it should be noted that the present disclosure can pre-train a service processing model, and the trained service processing model can include a central model and multiple auxiliary models. The above-mentioned central model is used to determine the user's intention based on the service demand information after receiving the user's service demand information input by the operation and maintenance personnel, and split the user's intention into at least one detailed intention, and then determine the target auxiliary model to be called according to the detailed intention, call the above-mentioned target auxiliary model and process the business demand data returned by the target auxiliary model to obtain the final service processing result. The above-mentioned auxiliary model is used to feedback the business demand data associated with the detailed intent after receiving the detailed intent sent by the central model.

[0073] refer to Figure 2 , Figure 2 A flowchart showing how to obtain a trained service processing model in an embodiment of the present disclosure includes steps S201 to S202:

[0074] In step S201, a trained central model and multiple trained auxiliary models are obtained.

[0075] In this step, illustratively, the above-mentioned trained central model and the trained multiple auxiliary models can all be trained using a reinforcement learning algorithm.

[0076] refer to Figure 3 , Figure 3 A flowchart showing how to obtain the trained central model in the embodiment of the present disclosure includes steps S301 to S304:

[0077] In step S301, a first training sample is collected.

[0078] In this step, the first training samples summarized and sorted by the billing operation and maintenance personnel may be collected. The first training samples may include service demand information samples of sample users, service processing result samples for the service demand information samples, and business data samples of sample users for multiple business sectors.

[0079] Among them, sample users refer to specific user groups that participate in the data collection or model training process. These users provide their behaviors, preferences, interaction data, etc. for analysis and modeling.

[0080] The service demand information sample may be a demand proposed by a sample user or information after a combination of multiple demands, such as one or more of: billing questions, billing rule questions, broadband fault problem consultation, etc.

[0081] The above-mentioned service processing result samples are the query or processing results fed back for the above-mentioned service demand samples, for example: the question answering results for the above-mentioned bill questions, the rule answering results for the above-mentioned billing rule questions, the problem troubleshooting results for the above-mentioned broadband failure problem consultation, etc.

[0082] The above-mentioned multiple business sections (also referred to as business categories) may be a billing section, an accounting section, a recharge section, a strategy section, a settlement section, etc. New sections may be added or existing sections may be removed according to actual business conditions, and the present disclosure does not make any special limitations on this.

[0083] The business data samples associated with the above-mentioned multiple business segments may be relevant activity records and data of sample users in different business segments, such as: package fees of sample users (for billing segment), recharge records of sample users (for recharge segment), etc., which can be set according to actual conditions, and this disclosure does not make any special limitations on this.

[0084] In step S302, a first training sample is input into a central model to be trained, and a first action strategy for the service demand information sample is output through the central model to be trained.

[0085] In this step, the above-mentioned first training sample can be input into the central model to be trained, and the first action strategy for the above-mentioned service demand information sample can be output through the above-mentioned central model to be trained, wherein the above-mentioned first action strategy is selected from a predefined action space range (a series of actions that can be executed by a pre-specified model, such as: triggering executable code, generating consulting results, triggering alarms, etc.), and is used to characterize the decision-making process of how to obtain a service processing result sample for the service demand information sample.

[0086] In step S303 , a first reward value for the first action strategy is determined according to a preconfigured first reward function.

[0087] In this step, after the central model to be trained outputs the first action strategy, the first reward value for the first action strategy can be determined according to a preset reward function (which may include multiple reward and punishment conditions, used to judge whether to reward or punish the first action strategy, as well as the reward value or penalty value when rewarding or punishing).

[0088] In step S304, when the first reward value reaches the first training termination condition, the training of the central model to be trained is terminated to obtain a trained central model.

[0089] In this step, when the first reward value reaches the first training termination condition, for example, the first reward value tends to converge or maximize (for example, greater than the preset reward value), the training of the central model to be trained can be stopped to obtain the trained central model.

[0090] refer to Figure 4 , Figure 4 A flowchart showing how to obtain each trained auxiliary model in the embodiment of the present disclosure includes steps S401 to S404:

[0091] In step S401, a second training sample is collected.

[0092] In this step, a second training sample can be collected, which includes a refined intention sample of the sample user (which can be a sub-intention after the user intention corresponding to the above-mentioned service demand sample is split), a business demand data sample for the refined intention sample (for example: when the refined intention is a consumption record query, the business demand data sample can be the consumption record query result of the sample user) and a business data sample of the sample user for each business segment (for example: when training an auxiliary model related to the recharge segment, it can be the recharge record of the sample user).

[0093] In step S402, the second training sample is input into each auxiliary model to be trained, and each auxiliary model to be trained outputs a second action strategy for the refined intention sample.

[0094] In this step, the second training sample can be input into each auxiliary model to be trained, and each auxiliary model to be trained outputs a second action strategy for the refined intent sample. The second action strategy is selected from a predefined action space range and is used to characterize the decision-making process of how to obtain a business demand data sample for the refined intent sample.

[0095] In step S403, a second reward value for the second action strategy is determined according to a preconfigured second reward function.

[0096] In this step, after the central model to be trained outputs the second action strategy, the second reward value for the second action strategy can be determined according to a preset reward function (which may include multiple reward and punishment conditions, used to judge whether to reward or punish the second action strategy, as well as the reward value or penalty value when rewarding or punishing).

[0097] In step S404, when the second reward value reaches the second training termination condition, the training of each auxiliary model to be trained is terminated to obtain each trained auxiliary model.

[0098] In this step, when the second reward value reaches the second training termination condition, for example, the first reward value tends to converge or maximize (for example, greater than the preset reward value), the training of each auxiliary model to be trained can be stopped to obtain each trained auxiliary model.

[0099] After obtaining the trained central model and multiple trained auxiliary models, the process may proceed to step S202 to jointly train the trained central model and multiple trained auxiliary models to obtain a trained service processing model.

[0100] In this step, in order to ensure that the central model can better collaborate with multiple auxiliary models to perform business processing, the trained central model and multiple trained auxiliary models can be jointly trained to obtain a trained service processing model.

[0101] It should be noted that after obtaining the trained service processing model, the present invention can also use the actual business data within a preset time period to update the training samples in real time, and then use the updated training samples to optimize the trained service processing model again, so as to continuously optimize the service processing model and improve its processing performance.

[0102] refer to Figure 5 , Figure 5 A schematic diagram showing the structure of the trained service processing model in the embodiment of the present disclosure is shown as follows: Figure 5 As shown:

[0103] The above-mentioned central model can also be called the billing query and troubleshooting agent, and its function is to conduct multimodal interaction with the billing operation and maintenance personnel to receive the user's service demand information input by the billing operation and maintenance personnel, and determine the user's intention based on the service demand information, and decide how to call the target auxiliary model based on the user's intention, so as to determine the service processing result based on the call information fed back by the target auxiliary model.

[0104] The above-mentioned multiple auxiliary models may include: billing agent, account agent, recharge agent, strategy agent and settlement agent; among which different auxiliary models are associated with different business segments, and each business segment may contain multiple sub-business segments.

[0105] The business section associated with the above-mentioned billing agent may be a billing section, and the multiple sub-business sections contained in the billing section may be online billing and offline billing. Each sub-business section is associated with a sub-model, which may be an online billing sub-model, an offline billing sub-model, etc.;

[0106] The business segment associated with the above-mentioned accounting agent may be an accounting segment, and the multiple sub-business segments contained in the accounting segment may be bills, balances, invoices, write-offs, etc. Each sub-business segment is associated with a sub-model, which may be a bill sub-model, a balance sub-model, an invoice sub-model, a write-off sub-model, etc.;

[0107] The business section associated with the above-mentioned recharge agent may be a recharge section, and the multiple sub-business sections contained in the recharge section may be card recharge, non-card recharge, card management, reconciliation, etc. Each sub-business section is associated with a sub-model, which may be a card recharge sub-model, a non-card recharge sub-model, a card management sub-model, a reconciliation sub-model, etc.;

[0108] The business segment associated with the above-mentioned strategic agent may be a strategy segment, and the multiple sub-business segments associated with the strategy segment may be signal control, reminder, 4G strategy, 5G strategy, etc. Each sub-business segment is associated with a sub-model, which may be a signal control sub-model, a reminder sub-model, a 4G strategy sub-model, a 5G strategy sub-model, etc.;

[0109] The business segment associated with the above settlement agent may be a settlement segment, and the multiple sub-business segments associated with the settlement segment may be operator settlement, commission settlement, commission payment, account adjustment, etc. Each sub-business segment is associated with a sub-model, which may be an operator settlement sub-model, a commission settlement sub-model, a commission payment sub-model, an account adjustment sub-model, etc.;

[0110] In addition, it should be noted that each of the above-mentioned intelligent agents has at least the following core capabilities:

[0111] Intent recognition (understanding the purpose or need of user input, for example, when the user says "I want to know my account balance", the system should be able to recognize the user's intention to check the account balance), model management (for example: operations involving model deployment, updating, version control, and performance monitoring to ensure that the model used is the latest and has the best performance), session management (responsible for tracking and maintaining the history of conversations, understanding the context to ensure conversation coherence and consistency, for example, keeping the user's request and previously provided information in mind in multiple rounds of conversations), prompt engineering (used to design effective prompt statements to guide AI to generate the required response, which is particularly important in natural language processing tasks such as text generation or chatbots), data management (including a series of activities such as data collection, storage, cleaning, conversion, and analysis. Ensuring data quality and availability is crucial for training machine learning models), model fine-tuning (based on the pre-trained model, using data from a specific field to further train the model to better adapt to specific tasks or application scenarios), multi-round conversations (supporting multiple interactions to complete a task or solve a problem), plug-in management (allowing the system's functionality to be extended by adding or removing plug-ins. These plug-ins can provide additional services or the ability to integrate other third-party services) and other core capabilities.

[0112] Exemplarily, the interaction process between the above-mentioned central model and each auxiliary model can adopt a predefined fixed topology (that is, the central model is responsible for calling each auxiliary model and collecting the business query results fed back by each auxiliary model, and then feeding back the results to the billing personnel in an interactive mode).

[0113] Since the sub-business blocks contained in each business block can be added or deleted as the business changes, the number of sub-business blocks associated with each business block may change, and thus the number of sub-models contained in each auxiliary model may also change. Therefore, a learnable information interaction structure can be adopted between sub-models. For example, a control unit can be set between any two sub-models (sub-models contain policy / value networks), refer to Figure 6 , Figure 6 The structure diagram of the control unit in the embodiment of the present disclosure is shown as follows: Figure 6 As shown, the control unit can control the timing and object of information interaction based on the historical interaction between the two sub-models (for example, the generation or reception time of historical messages) to reduce the cost of information interaction. The control unit will refer to historical data when determining the priority of messages. If a certain type of message has had a significant positive impact on the processing of a certain type of decision-making scenario in the past, then this type of message will be given a higher priority when processing similar scenarios in the future. Therefore, when adding or deleting sub-models according to business needs, the control unit can adjust or reselect the interaction objects accordingly through learning.

[0114] After obtaining the above trained service processing model, you can then refer to Figure 1 , in step S110, after receiving the user's service demand information, the central model determines the user's intention based on the service demand information.

[0115] In this step, the user's service demand information may be consultation, fault reporting, complaint and other information related to billing, which may be specifically manifested as fee disputes, information control shutdown and resumption of the machine, bills and invoices, etc. Exemplarily, the service demand information may include at least one of consumption inquiry demand and troubleshooting demand.

[0116] Optionally, the service demand information in the present disclosure may be some complex user demands that require analysis, which cannot be directly answered by the account manager or customer service and need to be transferred to the second-line billing operation and maintenance personnel for processing. Therefore, the second-line operation and maintenance personnel can input the language description corresponding to the above service demand information into the above-trained service processing model, so that the trained service processing model receives the above service demand information.

[0117] Afterwards, the central model can determine the user intent based on the service demand information. The user intent is the user's purpose, desire or expected result. Exemplarily, the central model can perform intent recognition on the service demand information to determine the user intent.

[0118] For example, if the service demand information is: the user wants to check the phone balance and the recharge record of last month, then the user intent determined by the central model can be: check the phone balance; check the recharge record of last month.

[0119] In step S120, the central model determines the target auxiliary model to be called according to the user intention, calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model.

[0120] In this step, after determining the user intent, the central model can split the user intent to obtain at least one refined intent after the split; then, according to the target business segment associated with each refined intent, determine the target auxiliary model to be called.

[0121] For example, the first refined intention after splitting can be: query the phone bill balance, and the second refined intention can be: query the recharge record of last month, so it can be determined that the business segment associated with the first refined intention is the billing segment, so the auxiliary model corresponding to the billing segment can be determined as the first target auxiliary model to be called, and it can be determined that the business segment associated with the second refined intention is the recharge segment, so the auxiliary model corresponding to the recharge segment can be determined as the second target auxiliary model to be called.

[0122] When there are multiple target auxiliary models to be called, refer to Figure 7 , Figure 7 A flow chart showing how the central model calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model in the embodiment of the present disclosure includes steps S701 to S703:

[0123] In step S701, the central model decides the order of calling multiple target auxiliary models.

[0124] In this step, the central model may decide the order of calling the above-mentioned multiple target auxiliary models randomly or according to a predefined interaction order.

[0125] In step S702, each target auxiliary model is called in a calling order, and each target auxiliary model calls at least one target sub-business segment associated with each detailed intent to obtain business demand data for each detailed intent.

[0126] In this step, the central model can call each target auxiliary model in the above-mentioned calling order, and each target auxiliary model can call the target sub-model associated with at least one target sub-business segment under the target business segment according to each refined intent, to obtain the business demand data for each refined intent, for example: the balance query data for the first refined intent "phone bill balance", and the recharge record query data for the second refined intent "last month's recharge record".

[0127] In step S703, the business demand data is fed back to the central model through each target auxiliary model.

[0128] In this step, after each target auxiliary model has queried the corresponding business demand data, the business demand data can be actively fed back to the above-mentioned central model.

[0129] Next reference Figure 1 In step S130, the central model processes the business demand data to obtain a service processing result.

[0130] In this step, in an optional implementation, when the number of target auxiliary models called is one, the central model can process (for example, polish) the business demand data fed back by the target auxiliary model to obtain a service processing result.

[0131] When the number of target auxiliary models called is multiple, you can refer to Figure 8 , Figure 8A flowchart showing how to obtain a service processing result in an embodiment of the present disclosure includes steps S801 to S803:

[0132] In step S801, multiple business demand data are integrated to obtain preliminary processing results.

[0133] In this step, the central model can integrate the business demand data fed back by multiple target auxiliary models (for example, arranging multiple business demand data into one piece of data) to obtain preliminary processing results.

[0134] In step S802, key information is extracted from the preliminary processing result to obtain an intermediate processing result.

[0135] In this step, the central model can extract key information from the preliminary processing results to obtain intermediate processing results. Specifically, the intermediate processing results can be obtained by matching numbers and keywords from the preliminary processing results with regular expressions, or by identifying specific entities (e.g., key information such as amount, date, etc.) from the preliminary processing results based on a trained information extraction model to obtain the intermediate processing results.

[0136] In step S803, the intermediate processing result is reorganized according to the pre-learned template speech to obtain the service processing result.

[0137] In this step, after obtaining the intermediate processing results, the intermediate processing results can be reorganized according to the pre-learned template words to obtain the service processing results. For example, assuming that for "phone charge balance", the pre-defined template words are: Your current phone charge balance is {balance} yuan. Then the intermediate processing results can be nested into the template words to generate the service processing results.

[0138] It should be noted that, given that the above-mentioned service demand information may include consumption query demand and troubleshooting demand, the service processing result obtained in the present disclosure may include at least one of the consumption query result and the troubleshooting result. After obtaining the service processing result, the central model may report the above-mentioned service processing result to the operation and maintenance personnel, so that the operation and maintenance personnel can feedback the service processing result to the user.

[0139] The following describes the implementation methods of the present disclosure in combination with specific application scenarios:

[0140] After the central model (which can be called the billing query troubleshooting agent) and multiple auxiliary models (models can be called agents) have been trained and fine-tuned, and various executable codes have been installed as plug-ins, the billing operation and maintenance personnel can use natural language to describe the user's service demand information. After that, the billing query troubleshooting agent can determine the user's specific intention, taking the user's intention as an example of discount details analysis;

[0141] The billing query troubleshooting agent can split the user's intent, plan according to the segmented intent, make decisions on the target auxiliary models (for example, the accounting agent and the policy agent) that need to be called, and interact with the target auxiliary models;

[0142] The accounting agent determines and decides the sub-business segments that need to be called based on the segmentation intention (which can be the previous month's bills and discount details, the current month's bills and discount details);

[0143] The accounting agent calls the previous month's bill and discount details, the current month's bill and discount details, and returns the business demand data;

[0144] The strategic agent determines and decides the sub-business segments that need to be called based on the segmented intent (which can be the user profile query API (Application Programming Interface));

[0145] The policy agent calls the user profile query API and returns the business requirement data;

[0146] The billing query troubleshooting model analyzes the business demand data returned by the accounting agent and the policy agent, and then polishes the data according to the template script and feeds it back to the billing operation and maintenance personnel.

[0147] refer to Fig. 9 , Fig. 9 A flow chart showing a service processing method in an embodiment of the present disclosure for a specific application scenario is shown, including steps S901 to S906:

[0148] In step S901, the user intention is determined based on the service demand information. For example, the service demand information may be "Why is the bill discount for this number XXXXXXX less than in previous months?"

[0149] In step S902, path planning is performed (i.e., the target auxiliary model to be called is determined, the above-mentioned account agent and strategy agent);

[0150] In step S903, the accounting agent calls the previous month's bill and discount details;

[0151] In step S904, the accounting agent calls the bill and discount details of this month;

[0152] In step S905, the policy agent calls the user profile query API;

[0153] In step S906, the business demand data fed back by the accounting agent and the strategy agent is received, summarized and refined, and fed back to the user. For example, a group discount optional package of the user expired last month, so the discount has become less than in previous months.

[0154] Based on the above technical solutions, the present invention creatively proposes a billing query system based on multi-level intelligent agents. The system makes full use of the excellent generalization ability of big models in general task processing, as well as the high efficiency of automatic planning, decision-making and execution of complex tasks. By innovatively combining big models and AI intelligent agent technology, the traditional manual query and troubleshooting processes are intelligently reconstructed, significantly improving the efficiency and accuracy of billing queries.

[0155] The present disclosure also provides a service processing device, wherein a trained service processing model is configured on the device, wherein the trained service processing model includes a central model and multiple auxiliary models, wherein the multiple auxiliary models are associated with multiple business segments, and different auxiliary models are associated with different business segments. Fig.10 A schematic diagram showing the structure of a service processing device in an exemplary embodiment of the present disclosure is shown; Fig.10 As shown, the service processing device 1000 may include a user intention determination module 1010, an auxiliary model calling module 1020 and a processing result determination module 1030. Among them:

[0156] A user intention determination module 1010 is used to determine the user intention according to the service demand information through the central model after receiving the user's service demand information;

[0157] An auxiliary model calling module 1020 is used to determine a target auxiliary model to be called according to the user intention through the central model, so that the target auxiliary model feeds back business demand data associated with the user intention;

[0158] The processing result determination module 1030 is used to process the business demand data through the central model to obtain a service processing result.

[0159] In an exemplary embodiment of the present disclosure, the user intention determination module 1010 is configured to:

[0160] Splitting the user intent through the central model to obtain at least one refined intent after the splitting;

[0161] The target auxiliary model to be called is determined according to the target business segment associated with each of the refined intentions.

[0162] In an exemplary embodiment of the present disclosure, each of the business segments includes a plurality of sub-business segments, and each of the sub-business segments is associated with a sub-model;

[0163] When there are multiple target auxiliary models to be called, the auxiliary model calling module 1020 is configured as follows:

[0164] Determine the calling order of the plurality of target auxiliary models through the central model;

[0165] Calling each of the target auxiliary models in the calling order, and calling at least one target sub-model associated with a target sub-business segment according to each of the detailed intents through each of the target auxiliary models to obtain business demand data for each of the detailed intents;

[0166] The business demand data is fed back to the central model through each of the target auxiliary models.

[0167] In an exemplary embodiment of the present disclosure, the processing result determination module 1030 processes the business demand data to obtain a service processing result, including:

[0168] Integrate the plurality of business demand data to obtain a preliminary processing result;

[0169] Extracting key information from the preliminary processing results to obtain intermediate processing results;

[0170] The intermediate processing result is reorganized according to the pre-learned template speech to obtain the service processing result.

[0171] In an exemplary embodiment of the present disclosure, after obtaining the service processing result, the processing result determination module 1030 is configured to:

[0172] The central model reports the service processing result to the operation and maintenance personnel, so that the operation and maintenance personnel feed back the service processing result to the user.

[0173] In an exemplary embodiment of the present disclosure, the trained service processing model is obtained by:

[0174] Obtain a trained central model and multiple trained auxiliary models;

[0175] The trained central model and multiple trained auxiliary models are jointly trained to obtain the trained service processing model.

[0176] In an exemplary embodiment of the present disclosure, the trained central model is obtained by training in the following manner:

[0177] Collecting a first training sample; the first training sample includes a service demand information sample of a sample user, a service processing result sample for the service demand information sample, and a business data sample of the sample user for the multiple business sectors;

[0178] Inputting the first training sample into the central model to be trained, and outputting a first action strategy for the service demand information sample through the central model to be trained; the first action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the service processing result sample for the service demand information sample;

[0179] Determine a first reward value for the first action strategy according to a preconfigured first reward function;

[0180] When the first reward value reaches a first training termination condition, the training of the central model to be trained is terminated to obtain the trained central model.

[0181] In an exemplary embodiment of the present disclosure, each trained auxiliary model is trained in the following manner:

[0182] Collecting a second training sample; the second training sample includes a refined intention sample of a sample user, a business demand data sample for the refined intention sample, and a business data sample of the sample user for each of the business sectors;

[0183] The second training sample is input into each auxiliary model to be trained, and each auxiliary model to be trained outputs a second action strategy for the refined intent sample; the second action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the business demand data sample for the refined intent sample;

[0184] Determine a second reward value for the second action strategy according to a preconfigured second reward function;

[0185] When the second reward value reaches the second training termination condition, the training of each auxiliary model to be trained is terminated to obtain each trained auxiliary model.

[0186] The specific details of each module in the above service processing device have been described in detail in the corresponding service processing method, so they will not be repeated here.

[0187] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0188] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0189] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0190] The present disclosure also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or may exist independently without being assembled into the electronic device.

[0191] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0192] Computer-readable storage media can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable storage medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0193] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by an electronic device, the electronic device implements the method described in the above embodiments.

[0194] In addition, an electronic device capable of implementing the above method is also provided in an embodiment of the present disclosure.

[0195] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0196] Refer to the following Fig.11 1100 according to this embodiment of the present disclosure is described. Fig.11 The electronic device 1100 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0197] like Fig.11 As shown, the electronic device 1100 is in the form of a general computing device. The components of the electronic device 1100 may include, but are not limited to: at least one processor 1110, at least one memory 1120, a bus 1130 connecting different system components (including the memory 1120 and the processor 1110), and a display 1140.

[0198] The memory stores program codes, which can be executed by the processor 1110, so that the processor 1110 executes the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processor 1110 can execute the following steps: Figure 1 As shown in: Step S110, after receiving the user's service demand information, the central model determines the user's intention based on the service demand information; Step S120, the central model determines the target auxiliary model to be called according to the user's intention, calls the target auxiliary model and obtains the business demand data fed back by the target auxiliary model; Step S130, the central model processes the business demand data to obtain the service processing result.

[0199] The memory 1120 may include a readable medium in the form of a volatile storage, such as a random access memory (RAM) 11201 and / or a cache memory 11202 , and may further include a read-only memory (ROM) 11203 .

[0200] Memory 1120 may also include a program / utility 11204 having a set (at least one) of program modules 11205, such program modules 11205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0201] The bus 1130 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0202] The electronic device 1100 may also communicate with one or more external devices 1200 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1100, and / or communicate with any device that enables the electronic device 1100 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1150. Furthermore, the electronic device 1100 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1160. As shown, the network adapter 1160 communicates with other modules of the electronic device 1100 via a bus 1130. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1100, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0203] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A service processing method, characterized in that: Applied to a trained service processing model, the trained service processing model includes a central model and multiple auxiliary models, the multiple auxiliary models are associated with multiple business segments, and different auxiliary models are associated with different business segments, the method includes: After receiving the service demand information of the user, the central model determines the user intention according to the service demand information; The central model determines the target auxiliary model to be called according to the user intention, so that the target auxiliary model feeds back the business demand data associated with the user intention; The central model processes the business demand data to obtain a service processing result.

2. The method according to claim 1, characterized in that The central model determines the target auxiliary model to be called according to the user intention, including: The central model splits the user intent to obtain at least one refined intent after the split; The target auxiliary model to be called is determined according to the target business segment associated with each of the refined intentions.

3. The method according to claim 2, characterized in that Each of the business sections includes multiple sub-business sections, and each of the sub-business sections is associated with a sub-model; When there are multiple target auxiliary models to be called, calling the target auxiliary model and obtaining the business demand data fed back by the target auxiliary model includes: The central model decides the order of calling the plurality of target auxiliary models; Calling each of the target auxiliary models in the calling order, and calling at least one target sub-model associated with a target sub-business segment according to each of the detailed intents through each of the target auxiliary models to obtain business demand data for each of the detailed intents; The business demand data is fed back to the central model through each of the target auxiliary models.

4. The method according to claim 3, characterized in that The processing of the business demand data to obtain a service processing result includes: Integrate the plurality of business demand data to obtain a preliminary processing result; Extracting key information from the preliminary processing results to obtain intermediate processing results; The intermediate processing result is reorganized according to the pre-learned template speech to obtain the service processing result.

5. The method according to claim 4, characterized in that After obtaining the service processing result, the method further includes: The central model reports the service processing result to the operation and maintenance personnel, so that the operation and maintenance personnel feed back the service processing result to the user.

6. The method according to claim 1, characterized in that The trained service processing model is obtained in the following way: Obtain a trained central model and multiple trained auxiliary models; The trained central model and multiple trained auxiliary models are jointly trained to obtain the trained service processing model.

7. The method according to claim 6, characterized in that The trained central model is obtained by training in the following way: Collecting a first training sample; the first training sample includes a service demand information sample of a sample user, a service processing result sample for the service demand information sample, and a business data sample of the sample user for the multiple business sectors; Inputting the first training sample into a central model to be trained, and outputting a first action strategy for the service demand information sample through the central model to be trained; The first action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the service processing result sample for the service demand information sample; Determine a first reward value for the first action strategy according to a preconfigured first reward function; When the first reward value reaches a first training termination condition, the training of the central model to be trained is terminated to obtain the trained central model.

8. The method according to claim 6, characterized in that Each trained auxiliary model is trained in the following way: Collecting a second training sample; the second training sample includes a refined intention sample of a sample user, a business demand data sample for the refined intention sample, and a business data sample of the sample user for each of the business sectors; Inputting the second training sample into each auxiliary model to be trained, and outputting a second action strategy for the refined intention sample through each auxiliary model to be trained; The second action strategy is selected from a predefined action space range and is used to characterize a decision process of how to obtain the business demand data sample for the refined intention sample; Determine a second reward value for the second action strategy according to a preconfigured second reward function; When the second reward value reaches the second training termination condition, the training of each auxiliary model to be trained is terminated to obtain each trained auxiliary model.

9. A service processing device, characterized in that: The device is configured with a trained service processing model, the trained service processing model includes a central model and multiple auxiliary models, the multiple auxiliary models are associated with multiple business segments, and different auxiliary models are associated with different business segments, and the device includes: A user intention determination module is used to determine the user intention according to the service demand information through the central model after receiving the user's service demand information; An auxiliary model calling module, used to determine a target auxiliary model to be called according to the user intention through the central model, so that the target auxiliary model feeds back business demand data associated with the user intention; The processing result determination module is used to process the business demand data through the central model to obtain a service processing result.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the service processing method according to any one of claims 1 to 8 is implemented.

11. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to execute the service processing method according to any one of claims 1 to 8 by executing the executable instructions.

Images