Service dialogue model training method and device, electronic equipment and storage medium

Abstract

The invention provides a service dialogue model training method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining a historical service call record and user portrait data of a target service; carrying out call link segment interception on the historical service call record to obtain call segments of a plurality of call links; generating a plurality of user simulation scripts for each user setting type in the plurality of user setting types based on the user portrait data and call feature information obtained by analyzing call segments of the plurality of call links; and based on a plurality of user simulation scripts of a set type of each user, carrying out simulation service dialogue on the user artificial intelligence agent model and the consultant artificial intelligence agent model to obtain simulation service dialogue data, and training the initial service dialogue model by taking the simulation service dialogue data as training data to obtain a target service dialogue model. The labor cost in the business dialogue model training process can be reduced, and the dialogue response speed can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a business dialogue model training method and device, electronic equipment and storage medium. Background Art

[0002] Task-oriented dialogue systems (TOD) can be used to solve specific problems in a certain field, such as helping users find products, book movie tickets, make restaurant reservations, and conduct business consultation and transactions in professional fields.

[0003] When solving specific problems in a certain field based on TOD, it is necessary to first obtain rich language data in that field, and then train an end-to-end TOD robot based on the acquired language data. The TOD robot can understand the information conveyed by the user through language and feedback the information the user wants, thereby helping the user solve specific problems in a certain field.

[0004] At present, although historical conversation data can be obtained from business systems in some professional fields as training data for TOD robots, due to the uneven levels of salesmen in actual business consultation or answering, training TOD robots based only on historical conversation data obtained from business systems is often not professional and standardized. This requires adding a large amount of manually set annotated data to train TOD robots, so that TOD robots are more professional and standardized. At the same time, in order to ensure the accuracy and intelligence of TOD robot services, the number of model parameters of TOD robots is relatively large. In actual situations, manually setting a large amount of annotated data is time-consuming and labor-intensive, and requires a lot of labor costs; at the same time, the TOD robots trained based on the current training method have a large number of model parameters, a long response time to users, and a slow response speed to dialogue with users. Summary of the invention

[0005] The present invention provides a business dialogue model training method and device, an electronic device and a storage medium, which are used to solve the defects in the prior art that the training process of the business dialogue model requires a lot of labor costs and the response speed of the business dialogue model is slow, so as to achieve the purpose of reducing the labor costs required in the training process of the business dialogue model and improving the dialogue response speed, while ensuring the accuracy and intelligence of the business dialogue service process.

[0006] The present invention provides a business dialogue model training method, comprising the following steps.

[0007] Obtain historical business call records of the target business and user portrait data of the target business; intercept call segment fragments of the historical business call records to obtain call segments of multiple call segments; parse the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and call feature information, generate multiple user simulation scripts for each user setting type in multiple user setting types; input any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, simulate business conversations based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model, and obtain simulated business conversation data; wherein the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business; train the initial business conversation model based on the simulated business conversation data to obtain a target business conversation model; wherein the number of model parameters of the target business conversation model is less than the number of model parameters of the consultant artificial intelligence agent model.

[0008] According to a business dialogue model training method provided by the present invention, any user simulation script from multiple user simulation scripts of each user setting type is input into a preset user artificial intelligence agent model, and a simulated business dialogue is performed based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model to obtain simulated business dialogue data, including: when the current number of extractions changes and the change result is less than the extraction number threshold, any user simulation script is extracted from the multiple user simulation scripts of each user setting type as the user simulation script to be input, and a plurality of user simulation scripts to be input corresponding to multiple user setting types are obtained; the plurality of user simulation scripts to be input are input into the user artificial intelligence agent model to obtain a target user artificial intelligence agent model corresponding to the current number of extractions; based on the simulated business dialogue of the target user artificial intelligence agent model and the consultant artificial intelligence agent model, the business dialogue data corresponding to the current number of extractions is obtained; the current number of extractions is increased by 1; when the current number of extractions changes and the change result is greater than or equal to the extraction number threshold, the business dialogue data corresponding to all extraction times are determined as simulated business dialogue data.

[0009] According to a business dialogue model training method provided by the present invention, multiple user setting types include: user role personality setting, user call link end setting, and call link user intention setting; before inputting multiple user simulation scripts to be input into the user artificial intelligence agent model and obtaining the target user artificial intelligence agent model corresponding to the current number of extractions, the method also includes: obtaining a first user simulation script to be input corresponding to the user call link end setting and a second user simulation script to be input corresponding to the call link user intention setting from the multiple user simulation scripts to be input; analyzing whether the call link progress of the first user simulation script to be input and the second user simulation script to be input match; in the first user simulation script to be input When the call progress of the user simulation script and the second user simulation script to be input matches, the multiple user simulation scripts to be input are input into the user artificial intelligence agent model; when the call progress of the first user simulation script to be input and the second user simulation script to be input do not match, any user simulation script is re-extracted from the multiple user simulation scripts set at the end of the user's call phase as the first user simulation script to be input, or any user simulation script is re-extracted from the multiple user simulation scripts set according to the user's intention in the call phase as the second user simulation script to be input, and then the analysis of whether the first user simulation script to be input and the second user simulation script to be input match and its subsequent steps are re-executed.

[0010] According to a business dialogue model training method provided by the present invention, before inputting any user simulation script among multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, it also includes: receiving a custom script set for any target user setting type among multiple user setting types; using the custom script as a user simulation script of the target user setting type, and adding the custom script to the multiple user simulation scripts of the target user setting type.

[0011] According to a business dialogue model training method provided by the present invention, the target feature information also includes: user feature information; based on the user portrait data and call feature information, before generating multiple user simulation scripts for each user setting type in multiple user setting types, it also includes: generating user portrait completion data according to the user feature information; and completing the user portrait data according to the user portrait completion data.

[0012] According to a business dialogue model training method provided by the present invention, the multiple call links include: a call description link, a confirmation guidance link and a payment guidance link.

[0013] The present invention also provides a business dialogue model training device, which includes the following modules: an acquisition module, an interception module, a parsing module, a generation module, a processing module and a training module.

[0014] The acquisition module is used to obtain the historical business call records of the target business and the user portrait data of the target business.

[0015] The interception module is used to intercept call segment fragments from historical business call records to obtain call segments of multiple call segments.

[0016] The parsing module is used to parse the call segments of multiple call links to obtain target feature information; wherein the target feature information includes: call feature information.

[0017] A generation module is used to generate multiple user simulation scripts for each user setting type in multiple user setting types based on user portrait data and call feature information.

[0018] A processing module is used to input any user simulation script from multiple user simulation scripts of each user set type into a preset user artificial intelligence agent model, simulate business conversations based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model, and obtain simulated business conversation data; wherein, the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business.

[0019] A training module is used to train an initial business dialogue model based on simulated business dialogue data to obtain a target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the advisor artificial intelligence agent model.

[0020] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-mentioned business dialogue model training methods is implemented.

[0021] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned business dialogue model training methods.

[0022] The present invention also provides a computer program product, including a computer program, which implements any of the above-mentioned business dialogue model training methods when executed by a processor.

[0023] The business dialogue model training method and device, electronic device and storage medium provided by the present invention obtain historical business call records and user portrait data of the target business; intercept call segment fragments of the historical business call records to obtain call segments of multiple call segments; parse the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and the call feature information, generate multiple user simulation scripts for each user setting type in multiple user setting types; input any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, simulate business dialogue based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model, and obtain simulated business dialogue data; wherein the simulated business dialogue data includes: user dialogue data and consultant dialogue data, and the consultant artificial intelligence agent model generates consultant dialogue data based on multiple business databases of the target business; train the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model. It can be seen that in the process of obtaining the target business dialogue model, the present invention does not need to obtain a large amount of annotated data in a manual setting manner, but by setting multiple user setting types, generating multiple user simulation scripts for each user setting type, and then generating multiple user simulation scripts based on each user setting type, obtaining simulated business dialogue data based on multiple user simulation scripts of each user setting type, user artificial intelligence agent model and consultant artificial intelligence agent model, using the simulated business dialogue data as the training data of the initial business dialogue model to train the initial business dialogue model, and finally obtaining the target business dialogue model, and the number of model parameters of the target business dialogue model is equal to the number of model parameters of the consultant artificial intelligence agent model. Therefore, the present invention can solve the defects of the prior art that the training process of the business dialogue model requires a lot of labor costs and the response speed of the business dialogue model is slow, and achieves the purpose of reducing the labor costs to be invested in the training process of the business dialogue model and improving the dialogue response speed, while ensuring the accuracy and intelligence of the business dialogue service process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1This is one of the flow charts of the business dialogue model training method provided by the present invention.

[0026] Figure 2 It is a schematic diagram of capturing call segment segments from historical business call records in the business conversation model training method provided by the present invention.

[0027] Figure 3 It is a schematic diagram of parsing call segments of multiple call links in the business conversation model training method provided by the present invention.

[0028] Figure 4 It is a schematic diagram of the consultant artificial intelligence agent model generating consultant dialogue data in the business dialogue model training method provided by the present invention.

[0029] Figure 5 This is the second flow chart of the business dialogue model training method provided by the present invention.

[0030] Figure 6 It is a schematic diagram of the process of simulated dialogue between the target user artificial intelligence agent model and the consultant artificial intelligence agent model in the business dialogue model training method provided by the present invention.

[0031] Figure 7 This is the third flow chart of the business dialogue model training method provided by the present invention.

[0032] Figure 8 This is the fourth flow chart of the business dialogue model training method provided by the present invention.

[0033] Fig. 9 It is a working schematic diagram of the target business dialogue model in the business dialogue model training method provided by the present invention.

[0034] Fig.10 It is a structural schematic diagram of the business dialogue model training device provided by the present invention.

[0035] Fig.11 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Combine the following Figure 1-Figure 9 The business dialogue model training method of the present invention is described.

[0038] Figure 1 This is one of the flow charts of the business dialogue model training method provided by the present invention. Figure 1 As shown, the method includes the following S110~S160.

[0039] S110: Obtain historical service call records of the target service and user profile data of the target service.

[0040] The target business may be a business related to mass consumption, such as tourism business, catering business, ticket booking business, and so on.

[0041] The target business can also be business in some professional fields, such as insurance business, financial business, legal consulting business, etc.

[0042] The historical business call records of the target business may be business call records between the salesperson and the user in history obtained from the business system of the target business. The business call records may include, for example, business consultation, business description, business marketing, business suggestions, business return visits and many other business-related call records.

[0043] The user portrait data of the target business can be obtained in the following ways. First, the basic information data of multiple users of the target business can be obtained from the business system of the target business. The basic information data of each user can include, for example, age, gender, education, occupation, income, and other information that can be used to describe user characteristics. After that, the basic information data of multiple users are cleaned, clustered, analyzed, and the like, and the basic information data of multiple users are abstracted into multiple user tags, such as age tags, gender tags, education tags, occupation tags, income tags, etc. Based on the above multiple user tags, multiple user portraits are reconstructed to obtain user portrait data. The user portrait data may include portrait data of multiple user portraits, and the portrait data of each user portrait may include n (n is a natural number and n is greater than 1) user tags, and the n user tags are used to describe the user characteristics of the user portrait.

[0044] S120: intercepting call segment fragments from historical business call records to obtain call segments of multiple call segments.

[0045] In actual business calls, the communication process between the salesperson and the user can usually be divided into multiple links (i.e., call links). For example, if the salesperson wants to recommend a certain business service to the user, after the call is connected, the salesperson must first introduce himself to the user and explain the reason for the call. This is a necessary call link, and we can define this call link as the call explanation link. When the user chooses to continue answering the call after learning the reason for the call, the salesperson can start the next call link, which is usually to recommend a certain business service, such as recommending insurance business. This process is usually accompanied by confirmation operations such as SMS confirmation and key confirmation to guide the user to confirm the intention to purchase the recommended service. We can define this call link as the confirmation guidance link; after the user completes all confirmation operations, it can be confirmed that the user wants to purchase the service, and the salesperson will explain to the user how to purchase and pay for the service. For example, the salesperson guides the user to enter the business payment page and guides the user to complete the payment operation, thereby completing the process of purchasing and paying for the service. We can define this call link as the payment guidance link.

[0046] It can be seen that in actual business calls, multiple call links between the salesperson and the user may usually include: call instructions link, confirmation guidance link and payment guidance link.

[0047] It is understandable that, in specific implementation, the name of each call link can be adjusted accordingly according to the actual business. For example, for the recommendation of insurance business, after the call explanation is completed to the user, if the user is interested in purchasing the recommended insurance business, the user needs to send a text message to the business system of the insurance business, or the business system of the insurance business will send a text message to the user, and the user needs to reply to the text message sent by the business system. Since the operations in the confirmation guidance link are all operations on text messages, the confirmation guidance link can be called the text message guidance link.

[0048] It is understandable that, in specific implementation, each call link can also set up multiple different sub-links according to the actual business. For example, for the recommended business of insurance business, in the payment guidance link, it is usually necessary to first guide the user to enter the payment page by performing relevant page jump operations, and then guide the user to complete payment-related operations on the payment page, and finally complete the entire payment process. Based on this, two sub-links can be set for the payment guidance link: the payment page entry guidance sub-link and the payment operation guidance sub-link. The payment page entry guidance sub-link is used to guide the user to enter the payment page by performing relevant page jump operations, and the payment operation guidance sub-link is used to guide the user to complete payment-related operations on the payment page.

[0049] For multiple pre-set call links, call link fragments can be extracted from historical business call records based on the preset first large language model (LargeLanguage Models, LLMs). For example, the fragments corresponding to each call link in the historical business call records can be first identified, and then the fragments corresponding to each call link can be split or extracted, so as to extract the fragments corresponding to each call link in the historical business call records, and finally obtain call fragments of multiple call links.

[0050] For example, see Figure 2 , the target business is the recommended business of insurance business, the historical business call records are input into the first LLMs, the call link segments of the historical business call records are intercepted through the first LLMs, and the first LLMs outputs the intercepted call description (corresponding to the call description segment) call segment, SMS sending (corresponding to the confirmation guidance segment) call segment, payment page entry (corresponding to the payment page entry guidance sub-segment) call segment, and payment guidance (corresponding to the payment operation guidance sub-segment) call segment.

[0051] In a specific implementation, the first initial LLMs may be trained according to actual business requirements to obtain the above-mentioned first LLMs.

[0052] S130: Analyze the call segments of multiple call links to obtain target feature information; wherein the target feature information includes: call feature information.

[0053] The target feature information may at least include call feature information, which may include, for example, call performance behavior information, call content intention information, call script flow information, and call style information.

[0054] It should be noted that the call feature information is information about relevant features extracted from the call records on the user side, and the call feature information also represents the call features on the user side.

[0055] The call performance behavior information may include, for example, interested behavior, calm behavior, impatient behavior, indifferent behavior, and other information related to the emotional performance behavior of the user during the call.

[0056] The call content intention information may include, for example, demand information, question information, objection information, and other information related to the call content intention of the user in the call.

[0057] The call script flow information may include, for example, the call link reached at the end of the call. For example, if the user hangs up the call during the call explanation link, the call script flow information may be the call explanation link; or if the user hangs up the call during the confirmation guidance link, the call script flow information may be the confirmation guidance link.

[0058] The call style information may include, for example, information related to the user's conversation style, such as talkative, easy-going, silent, cheerful, active, and aloof.

[0059] In some embodiments, the target feature information may also include: user feature information. The user feature information may specifically be information related to the user's background, which may be extracted from the historical business call records between the user and the salesperson, such as the city where the user lives, the user's consumption habits, etc. Of course, it is not excluded that the user's basic information data such as the user's age, gender, education, occupation, income, etc. may also be extracted from the historical business call records.

[0060] In a specific implementation, based on the preset second LLMs, the call segments of multiple call links can be parsed to obtain the above target feature information. For the acquisition process of the second LLMs, the second initial LLMs can be trained according to the actual business needs to obtain the above second LLMs.

[0061] For example, Figure 3 As shown, call segments of multiple call links are input into the second LLMs, the second LLMs parse the call segments of multiple call links, and then output user feature information, call performance behavior information, call content intention information, call script flow information, and call style information.

[0062] In some embodiments, when the target feature information also includes user feature information, before executing the subsequent step S140, see Figure 3 , user portrait completion data can also be generated according to the user characteristic information obtained above; and then the user portrait data can be completed according to the user portrait completion data.

[0063] Specifically, the process of generating user portrait completion data based on user feature information may include, for example: comparing the user feature information with all tags of each user portrait in the user portrait data, determining that the user feature information is similar to the target user portrait when the user feature information overlaps with a set proportion of tags in the target user portrait, obtaining target information in the user feature information that does not appear in all tags of the target user portrait, and determining the target information as the portrait completion data of the target user portrait. When completing the target user portrait, at least one portrait completion tag is generated based on the portrait completion data, and the at least one portrait completion tag is added to the multiple original tags of the target user portrait.

[0064] For example, user portrait 1 in the user portrait data has n user tags. By parsing call fragments of multiple call links, user feature information of a certain user (actually multiple user feature tags) is obtained, and the multiple user feature tags are compared with the n user tags of user portrait 1. When the comparison result is that the multiple user feature tags overlap with m (m is a natural number and m is less than or equal to n) user tags of user portrait 1, if m / n exceeds the set ratio, at least one user feature tag among the multiple user feature tags that does not appear in the n user tags of user portrait 11 is determined as at least one portrait completion tag, and the at least one portrait completion tag is added to the n user tags of user portrait 1.

[0065] S140: Based on the user portrait data and the call feature information, generate multiple user simulation scripts for each user setting type in the multiple user setting types.

[0066] The multiple user setting types may include, for example: user role personality setting, user call session end setting, and call session user intention setting.

[0067] Multiple user simulation scripts can be generated for user role personality settings based on user portrait data and call feature information. For example, the user portrait data includes portrait data of multiple user portraits, and the portrait data of multiple user portraits can be freely combined with call performance behavior information, call content intention information, call script process information, and call style information and other call feature information to generate multiple user simulation scripts for user role personality settings. In the multiple user simulation scripts for user role personality settings, the user role personality settings of each user simulation script are different. For example, each user simulation script may include at least: one or more of the basic user information such as age, gender, education, occupation, and income; and one or more of the information related to the user's conversation style such as talkative, easy-going, silent, cheerful, jumping, and high-cold; and one or more of the information related to the emotional expression behavior in the call such as interested behavior, calm behavior, impatient behavior, and indifferent behavior; and one or more of the information related to the call content intention such as appeal information, question information, and objection information; and the call link reached at the end of the call.

[0068] Similarly, multiple user simulation scripts can be generated for the end setting of the user's call link based on the user portrait data and call feature information. In the call feature information, the call script process information may include: the call link reached when the call ends. Therefore, multiple user simulation scripts are generated for the end setting of the user's call link, and the scenario content of the multiple user simulation scripts can be created based on the user portrait data and the call script process information in the call feature information. For example, based on the portrait data of user portrait 1 in the user portrait data, and the call link reached when the call ends is the call description link, a user simulation script with corresponding scenario content can be created; based on the portrait data of user portrait 2 in the user portrait data, and the call link reached when the call ends is the confirmation guidance link, a user simulation script with corresponding scenario content can be created; and so on, finally multiple user simulation scripts are generated for the end setting of the user's call link.

[0069] Of course, it is understandable that, in order to simulate as many different forms of call ending scenarios that may be encountered in reality as possible, when generating multiple user simulation scripts for the user call ending setting, the call feature information used to create the scenario content of the multiple user simulation scripts includes not only the call script process information, but also the call performance behavior information and the call style information. That is, when generating multiple user simulation scripts for the user call ending setting, the scenario content of the multiple user simulation scripts can be created based on the user portrait data and the call performance behavior information, the call script process information, and the call style information in the call feature information.

[0070] Similarly, multiple user simulation scripts can be generated for the user intent setting of the call link based on the user portrait data and call feature information. In the call feature information, the call content intent information may include, for example, information related to the call content intent, such as demand information, question information, and objection information. Therefore, multiple user simulation scripts are generated for the user intent setting of the call link, and the scenario content of the multiple user simulation scripts can be created based on the user portrait data and the call content intent information in the call feature information. For example, a user simulation script corresponding to the scenario content can be created based on the portrait data of user portrait 1 in the user portrait data and call content intent information 1; a user simulation script corresponding to the scenario content can be created based on the portrait data of user portrait 2 in the user portrait data and call content intent information 2; and so on, and finally multiple user simulation scripts are generated for the user intent setting of the call link.

[0071] Of course, it is understandable that in order to simulate as many different forms of user intention scenarios that may be encountered in reality as possible, when multiple user simulation scripts are generated for the user intention setting of the call phase, the call feature information used to create the scenario content of the multiple user simulation scripts includes not only the call content intention information, but also the call performance behavior information and the call style information. That is, when multiple user simulation scripts are generated for the user intention setting of the call phase, the scenario content of the multiple user simulation scripts can be created based on the user portrait data and the call performance behavior information, the call content intention information, and the call style information in the call feature information.

[0072] The multiple user setting types listed above are merely exemplary. In the embodiment of the present invention, the multiple user setting types include but are not limited to the user setting types listed above, and may also include: user call tendency settings. Call tendencies may include, for example, a tendency to continue a call, a tendency to end a call, a tendency to request other information, etc.

[0073] S150: Input any user simulation script from multiple user simulation scripts of each user set type into a preset user artificial intelligence agent model, simulate business conversations based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model, and obtain simulated business conversation data; wherein, the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business.

[0074] First, the user artificial intelligence agent model and the consultant artificial intelligence agent model are introduced.

[0075] Both the user artificial intelligence agent model and the consultant artificial intelligence agent model are AI (Artificial Intelligence) Agent.

[0076] For the user artificial intelligence agent model, the data input therein are multiple user simulation scripts to be input corresponding to multiple user setting types one by one (see the corresponding introduction in S520 in the following embodiment for details, which will not be elaborated here).

[0077] For the consultant artificial intelligence agent model, the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business.

[0078] Specifically, see Figure 4 As shown, the multiple business databases of the target business may include: a professional knowledge base, an action base, and a tool base.

[0079] The professional knowledge base stores professional knowledge related to the target business. For example, if the target business is insurance business, the professional knowledge base stores professional knowledge related to insurance business; for another example, if the target business is legal consulting business, the professional knowledge base stores professional knowledge related to law.

[0080] The action library is used to store action data of feedback actions for the user's performance behavior. The above feedback actions may include, for example: soothing actions, reminder actions, encouraging actions, etc. For example, if the user's performance behavior is impatience, the feedback action may be a soothing action; if the user's performance behavior is interested, the feedback action may be an encouraging action, etc. It can be understood that the feedback actions listed above are only exemplary, and the feedback actions of the action library in the embodiment of the present invention include but are not limited to the feedback actions listed above.

[0081] The tool library is used to store data related to business calculations. For example, if the target business is insurance business, the insurance premium corresponding to the insurance type that the user is interested in can be calculated for the user; if the target business is legal consulting business, the legal time limit information related to the event consulted by the user can be calculated based on the time information provided by the user and the information stipulated in the legal provisions related to the event consulted by the user.

[0082] See also Figure 4 ,Based on multiple business databases of the target business, the consultant AI agent model ,can make different combinations of relevant data in the professional ,knowledge base, action base, and tool base, and generate corresponding ,consultant dialogue data according to the information given by the user AI agent ,model.

[0083] In some embodiments, Figure 5 As shown, the execution process of S150 may include S510 to S560.

[0084] S510: Determine whether the current number of extractions is less than the extraction number threshold.

[0085] If the judgment result is yes, that is, the current number of extractions is less than the extraction number threshold, S520 is executed.

[0086] If the judgment result is no, that is, the current number of extractions is greater than or equal to the extraction number threshold, S560 is executed.

[0087] It should be noted that the number of extractions is initially 0. The extraction number threshold can be set by those skilled in the art according to actual conditions, and the embodiment of the present invention does not limit this.

[0088] S520: When the current number of extractions changes and the change result is less than the extraction number threshold, a user simulation script is randomly extracted from multiple user simulation scripts of each user setting type as the user simulation script to be input, and multiple user simulation scripts to be input corresponding to multiple user setting types are obtained.

[0089] Each user setting type corresponds to multiple user simulation scripts. In order to freely combine the user simulation scripts in the multiple user setting types, a user simulation script is randomly selected from the multiple user simulation scripts of each user setting type as the user simulation script to be input each time, and multiple user simulation scripts to be input corresponding to multiple user setting types are obtained. The multiple user simulation scripts to be input are used for combination in the subsequent steps (corresponding to S530). In the above manner, the user simulation scripts in the multiple user setting types can be freely combined to simulate all possible scenarios on the user side as much as possible.

[0090] S530: Input multiple user simulation scripts to be input into the user artificial intelligence agent model to obtain the target user artificial intelligence agent model corresponding to the current number of extractions.

[0091] Multiple user simulation scripts to be input can be directly input into the user artificial intelligence proxy model to obtain the target user artificial intelligence proxy model corresponding to the current number of extractions.

[0092] In some embodiments, when multiple user setting types are user role personality setting, user call phase end setting, and call phase user intention setting, it is also possible to detect whether there are call phase conflicts in multiple user simulation scripts to be input. For example, in the consultant simulation script to be input for the user call phase end setting, the user has already ended the call in the call description phase, but in the consultant simulation script to be input for the call phase user intention setting, the user also asked questions in the confirmation guidance phase, thereby causing a conflict. If a call phase conflict occurs, the user simulation script to be input is replaced. For details, please refer to the introduction in S710~S740 in the subsequent embodiments, which will not be repeated here.

[0093] S540: Based on the simulated business dialogue of the target user artificial intelligence agent model and the consultant artificial intelligence agent model, obtain the business dialogue data corresponding to the current number of extractions.

[0094] For example, the simulated business dialogue process of the target user artificial intelligence agent model and the consultant artificial intelligence agent model can be seen in Figure 6 Based on the question-answering process between the target user artificial intelligence agent model and the consultant artificial intelligence agent model, the corresponding business dialogue data can be obtained.

[0095] In specific implementations, the Prompt can also be set according to actual needs, and the Prompt can be input into the target user artificial intelligence agent model to guide the target user artificial intelligence agent model to generate text or instructions for specific output.

[0096] In the specific implementation, both the user AI agent model and the consultant AI agent model are LLMs. Therefore, in the process of simulating business dialogues based on the target user AI agent model and the consultant AI agent model, parameters such as temperature and Top-P can be adjusted to stimulate the language generation capabilities of the target user AI agent model and the consultant AI agent model.

[0097] The business conversation data may include user conversation data output by the target user artificial intelligence agent model, and consultant conversation data generated by the consultant artificial intelligence agent model for the user conversation data.

[0098] S550: Increase the current number of extractions by 1.

[0099] The current number of extractions is increased by 1, and then S510 is executed.

[0100] S560: Determine the business conversation data corresponding to all extraction times as simulated business conversation data.

[0101] In some embodiments, the multiple user setting types are user role personality setting, user call session end setting, and call session user intention setting. Figure 7 As shown, before executing S530, the following S710~S740 may also be executed.

[0102] S710: From a plurality of user simulation scripts to be input, obtain a first user simulation script to be input corresponding to the user call phase end setting and a second user simulation script to be input corresponding to the call phase user intention setting.

[0103] S720: Analyze whether the call progress of the first user simulation script to be input and the second user simulation script to be input match.

[0104] The call end in the first user simulation script to be input is set to the first call link, and the call end in the second user simulation script to be input is set to the second call link. If the first call link is before the second call link, it is determined that the call link progress of the first user simulation script to be input and the second user simulation script to be input do not match; if the first call link is the same as the second call link, or the first call link is after the second call link, it is determined that the call link progress of the first user simulation script to be input and the second user simulation script to be input match.

[0105] If the analysis result is yes, that is, the call progress of the first user simulation script to be input and the second user simulation script to be input match, execute S730; if the analysis result is no, that is, the call progress of the first user simulation script to be input and the second user simulation script to be input do not match, execute S740.

[0106] S730: Input multiple user simulation scripts to be input into the user artificial intelligence agent model.

[0107] S740: Re-extract any user simulation script from the multiple user simulation scripts set at the end of the user call phase as the first user simulation script to be input, or re-extract any user simulation script from the multiple user simulation scripts set by the user's intention during the call phase as the second user simulation script to be input.

[0108] After executing S740, S720 is executed again.

[0109] In some embodiments, in order to further improve the training data for training the initial business dialogue model (corresponding to the subsequent step S160), a custom script set for the user can also be input into the user artificial intelligence agent model. In this case, before executing S150, if Figure 8 As shown, the following S810~S820 can also be executed.

[0110] S810: Receive a custom script set for any target user setting type among a plurality of user setting types.

[0111] S820: Use the custom script as a user simulation script of the target user setting type, and add the custom script to multiple user simulation scripts of the target user setting type.

[0112] S160: Train the initial business dialogue model based on the simulated business dialogue data to obtain a target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model.

[0113] The initial business dialogue models may be LLMs, for example.

[0114] In the related art, the user artificial intelligence agent model and the consultant artificial intelligence agent model are TOD robots, and the number of their model parameters is usually about 145000000000, which is a large number. The response time to the user's dialogue data is long, and the response speed of the dialogue with the user is slow. The standard business dialogue model obtained by the business dialogue model training method according to the embodiment of the present invention has a number of model parameters of about 3000000000~7000000000, and the number of model parameters is greatly reduced, which effectively improves the response speed of the dialogue with the user. At the same time, in the embodiment of the present invention, the training data of the initial business dialogue model is simulated business dialogue data. Since the simulated business dialogue data is based on the user artificial intelligence agent model and the consultant artificial intelligence agent model After a large number of simulated dialogues, it has high accuracy and can ensure the accuracy and intelligence of the business dialogue service process. In addition, in the process of obtaining training data (simulated business conversation data), user portrait data and call feature information can be automatically obtained, and multiple user simulation scripts of each user setting type can be automatically generated based on the user portrait data and call feature information. Finally, the user simulation scripts of different user setting types are freely combined and input into the user artificial intelligence agent model, so that the user artificial intelligence agent model and the consultant artificial intelligence agent model can conduct business conversations. After a large number of iterations, training data for training the initial business conversation model is finally obtained. In the process of obtaining training data, a large amount of labeled data does not need to be set manually, so that a lot of labor costs can be saved, achieving the purpose of reducing the labor costs to be invested in the training process of the business conversation model and improving the conversation response speed, while ensuring the accuracy and intelligence of the business conversation service process.

[0115] In specific implementation, Fig. 9 As shown, the target service dialogue model in the embodiment of the present invention can be connected to the action library, professional knowledge library, and tool library of the target service. The target service dialogue model actively calls the user, analyzes the user's demands, questions, or objections based on the information fed back by the user on the phone, and then selects corresponding feedback data from the action library, professional knowledge library, and / or tool library, generates a feedback strategy based on the feedback data, feeds back corresponding information to the user based on the feedback strategy, and determines whether the call can be continued based on the user's response to the feedback strategy, for example Fig. 9 In the process, after explaining the incoming call to the user, if the user is interested based on the user's answer, the call can be advanced to the next stage, that is, SMS sending, and so on. The service process is made accurate and intelligent.

[0116] The business dialogue model training method provided by the present invention obtains historical business call records of the target business and user portrait data of the target business; intercepts call segment fragments of the historical business call records to obtain call segments of multiple call segments; parses the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and the call feature information, generates multiple user simulation scripts for each user setting type in multiple user setting types; inputs any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, simulates business dialogue based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model, and obtains simulated business dialogue data; wherein the simulated business dialogue data includes: user dialogue data and consultant dialogue data, and the consultant artificial intelligence agent model generates consultant dialogue data based on multiple business databases of the target business; trains the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model. It can be seen that in the process of obtaining the target business dialogue model, the present invention does not need to obtain a large amount of annotated data in a manual setting manner, but by setting multiple user setting types, generating multiple user simulation scripts for each user setting type, and then generating multiple user simulation scripts based on each user setting type, obtaining simulated business dialogue data based on multiple user simulation scripts of each user setting type, user artificial intelligence agent model and consultant artificial intelligence agent model, using the simulated business dialogue data as the training data of the initial business dialogue model to train the initial business dialogue model, and finally obtaining the target business dialogue model, the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model. Therefore, the present invention can solve the defects of the prior art that the training process of the business dialogue model requires a lot of labor costs and the response speed of the business dialogue model is slow, and achieves the purpose of reducing the labor costs to be invested in the training process of the business dialogue model and improving the dialogue response speed, while ensuring the accuracy and intelligence of the business dialogue service process.

[0117] The business dialogue model training device provided by the present invention is described below. The business dialogue model training device described below and the business dialogue model training method described above can be referenced to each other.

[0118] Fig.10 Schematic diagram of the structure of the business dialogue model training device provided by the present invention. Fig.10 As shown, the business dialogue model training device 1000 includes: an acquisition module 1001, an interception module 1002, a parsing module 1003, a generation module 1004, a processing module 1005 and a training module 1006.

[0119] The acquisition module 1001 is used to obtain the historical service call records of the target service and the user portrait data of the target service.

[0120] The interception module 1002 is used to intercept call segment fragments from historical business call records to obtain call segments of multiple call segments.

[0121] The analysis module 1003 is used to analyze the call segments of multiple call links to obtain target feature information; wherein the target feature information includes: call feature information.

[0122] The generating module 1004 is used to generate a plurality of user simulation scripts for each of a plurality of user setting types based on the user portrait data and the call feature information.

[0123] Processing module 1005 is used to input any user simulation script from multiple user simulation scripts of each user set type into a preset user artificial intelligence agent model, simulate business conversations based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model, and obtain simulated business conversation data; wherein, the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business.

[0124] The training module 1006 is used to train the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the advisor artificial intelligence agent model.

[0125] The business dialogue model training device provided by the present invention obtains historical business call records of the target business and user portrait data of the target business; intercepts call segment fragments of the historical business call records to obtain call segments of multiple call segments; parses the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and the call feature information, generates multiple user simulation scripts for each user setting type in multiple user setting types; inputs any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, simulates business dialogue based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model, and obtains simulated business dialogue data; wherein the simulated business dialogue data includes: user dialogue data and consultant dialogue data, and the consultant artificial intelligence agent model generates consultant dialogue data based on multiple business databases of the target business; trains the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model. It can be seen that in the process of obtaining the target business dialogue model, the present invention does not need to obtain a large amount of annotated data in a manual setting manner, but by setting multiple user setting types, generating multiple user simulation scripts for each user setting type, and then generating multiple user simulation scripts based on each user setting type, obtaining simulated business dialogue data based on multiple user simulation scripts of each user setting type, user artificial intelligence agent model and consultant artificial intelligence agent model, using the simulated business dialogue data as the training data of the initial business dialogue model to train the initial business dialogue model, and finally obtaining the target business dialogue model, the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model. Therefore, the present invention can solve the defects of the prior art that the training process of the business dialogue model requires a lot of labor costs and the response speed of the business dialogue model is slow, and achieves the purpose of reducing the labor costs to be invested in the training process of the business dialogue model and improving the dialogue response speed, while ensuring the accuracy and intelligence of the business dialogue service process.

[0126] Fig.11 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.11As shown, the electronic device may include: a processor (processor) 1110 , a communication interface (Communications Interface) 1120 , a memory (memory) 1130 and a communication bus 1140 , wherein the processor 1110 , the communication interface 1120 , and the memory 1130 communicate with each other via the communication bus 1140 . The processor 1110 can call the logic instructions in the memory 1130 to execute the business dialogue model training method, which includes: obtaining historical business call records of the target business and user portrait data of the target business; intercepting call segment fragments of the historical business call records to obtain call segments of multiple call segments; parsing the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and the call feature information, generating multiple user simulation scripts for each user setting type in multiple user setting types; inputting any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, and simulating business dialogue based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model to obtain simulated business dialogue data; wherein the simulated business dialogue data includes: user dialogue data and consultant dialogue data, and the consultant artificial intelligence agent model generates consultant dialogue data based on multiple business databases of the target business; training the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model.

[0127] In addition, the logic instructions in the above-mentioned memory 1130 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0128] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the business dialogue model training method provided by the above methods, which includes: obtaining historical business call records of the target business and user portrait data of the target business; intercepting call segment fragments of the historical business call records to obtain call segments of multiple call segments; parsing the call segments of multiple call segments to obtain target feature information; wherein the target feature information includes: call feature information; based on the user portrait data and the call feature information, for each of the multiple user setting types Generate multiple user simulation scripts based on the user setting type; input any user simulation script from the multiple user simulation scripts of each user setting type into the preset user artificial intelligence agent model, simulate business conversations based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model, and obtain simulated business conversation data; wherein, the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates consultant conversation data based on multiple business databases of the target business; train the initial business conversation model based on the simulated business conversation data to obtain the target business conversation model; wherein, the number of model parameters of the target business conversation model is less than the number of model parameters of the consultant artificial intelligence agent model.

[0129] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented by a processor to execute the business dialogue model training method provided by the above-mentioned methods when executed, the method comprising: obtaining historical business call records of the target business and user portrait data of the target business; intercepting call segment fragments of the historical business call records to obtain call segments of multiple call segments; parsing the call segments of multiple call segments to obtain target feature information; wherein the target feature information comprises: call feature information; based on the user portrait data and the call feature information, generating multiple user simulation scripts for each user setting type in multiple user setting types; inputting any user simulation script of the multiple user simulation scripts of each user setting type into a preset user artificial intelligence agent model, and simulating business dialogue based on the user artificial intelligence agent model and the preset consultant artificial intelligence agent model to obtain simulated business dialogue data; wherein the simulated business dialogue data comprises: user dialogue data and consultant dialogue data, and the consultant artificial intelligence agent model generates consultant dialogue data based on multiple business databases of the target business; training the initial business dialogue model based on the simulated business dialogue data to obtain the target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model.

[0130] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0131] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A business dialogue model training method, characterized in that: include: Obtain historical service call records of the target service and user profile data of the target service; Intercepting call segment fragments from the historical business call records to obtain call segments of multiple call segments; Parsing the call segments of the plurality of call links to obtain target feature information; wherein the target feature information includes: call feature information; Based on the user portrait data and the call feature information, generating multiple user simulation scripts for each user setting type in multiple user setting types; Input any one of the multiple user simulation scripts of each user-set type into a preset user artificial intelligence agent model, and perform a simulated business conversation based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model to obtain simulated business conversation data; wherein the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates the consultant conversation data based on multiple business databases of the target business; The initial business dialogue model is trained based on the simulated business dialogue data to obtain a target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model.

2. The business dialogue model training method according to claim 1, characterized in that: The step of inputting any one of the plurality of user simulation scripts of each user-set type into a preset user artificial intelligence agent model, performing a simulated business dialogue based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model, and obtaining simulated business dialogue data comprises: When the current number of extractions is changed and the result of the change is less than the extraction number threshold, a user simulation script is randomly extracted from the multiple user simulation scripts of each user setting type as the user simulation script to be input, and a plurality of user simulation scripts to be input corresponding to the multiple user setting types are obtained; Input the multiple user simulation scripts to be input into the user artificial intelligence agent model to obtain the target user artificial intelligence agent model corresponding to the current number of extractions; Based on the simulated business dialogue between the target user artificial intelligence agent model and the consultant artificial intelligence agent model, obtaining the business dialogue data corresponding to the current number of extractions; Add 1 to the current number of draws; When the current number of extractions is changed and the change result is greater than or equal to the number of extractions threshold, the business conversation data corresponding to all the number of extractions are determined as the simulated business conversation data.

3. The business dialogue model training method according to claim 2, characterized in that: The multiple user setting types include: user role personality setting, user call session end setting, and call session user intention setting; before inputting the multiple user simulation scripts to be input into the user artificial intelligence agent model to obtain the target user artificial intelligence agent model corresponding to the current number of extractions, the method further includes: From the multiple user simulation scripts to be input, obtain a first user simulation script to be input corresponding to the user call phase end setting and a second user simulation script to be input corresponding to the call phase user intention setting; Analyzing whether the call progress of the first user simulation script to be input and the second user simulation script to be input match; In the case where the call progress of the first user simulation script to be input and the second user simulation script to be input match, inputting the multiple user simulation scripts to be input into the user artificial intelligence agent model; In the case that the call progress of the first user simulation script to be input and the second user simulation script to be input do not match, any user simulation script is re-extracted from the multiple user simulation scripts set for the end of the user call phase as the first user simulation script to be input, or any user simulation script is re-extracted from the multiple user simulation scripts set by the user's intention in the call phase as the second user simulation script to be input, and then the analysis of whether the first user simulation script to be input and the second user simulation script to be input match and its subsequent steps are re-executed.

4. The business dialogue model training method according to any one of claims 1 to 3, characterized in that: Before inputting any one of the plurality of user simulation scripts of each user-set type into a preset user artificial intelligence agent model, the method further includes: receiving a custom script set for any target user setting type among the plurality of user setting types; The custom script is used as a user simulation script of the target user setting type, and the custom script is added to the multiple user simulation scripts of the target user setting type.

5. The business dialogue model training method according to claim 1, characterized in that: The target feature information also includes: user feature information; Before generating multiple user simulation scripts for each of the multiple user setting types based on the user portrait data and the call feature information, the method further includes: Generate user portrait completion data according to the user feature information; The user portrait data is completed according to the user portrait completion data.

6. The business dialogue model training method according to claim 1, characterized in that: The multiple call links include: a call description link, a confirmation guidance link and a payment guidance link.

7. A business dialogue model training device, characterized in that: include: An acquisition module, used to acquire historical service call records of a target service and user profile data of the target service; An interception module, used to intercept call segment fragments of the historical service call records to obtain call segments of multiple call segments; The parsing module is used to parse the call segments of the plurality of call links to obtain target feature information; wherein the target feature information includes: call feature information; A generating module, configured to generate a plurality of user simulation scripts for each of a plurality of user setting types based on the user portrait data and the call feature information; A processing module, used for inputting any user simulation script of the plurality of user simulation scripts of each user-set type into a preset user artificial intelligence agent model, performing a simulated business conversation based on the user artificial intelligence agent model and a preset consultant artificial intelligence agent model, and obtaining simulated business conversation data; wherein the simulated business conversation data includes: user conversation data and consultant conversation data, and the consultant artificial intelligence agent model generates the consultant conversation data based on a plurality of business databases of the target business; A training module is used to train an initial business dialogue model based on the simulated business dialogue data to obtain a target business dialogue model; wherein the number of model parameters of the target business dialogue model is less than the number of model parameters of the consultant artificial intelligence agent model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the business dialogue model training method as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the business dialogue model training method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the business dialogue model training method as described in any one of claims 1 to 6 is implemented.

Images