Recommended processing method and apparatus

Abstract

Embodiments of the present specification provide a recommendation processing method and device, wherein the recommendation processing method comprises: in the process of conducting a dialogue recommendation to a user through a multi-agent system when the user accesses a resource service, first conducting classification recommendation matching on dialogue data of the user through a scheduling agent, sending dialogue data that passes the matching to each classification agent, obtaining a detection result through task detection of a classification recommendation task on the dialogue data by each classification agent, inputting the detection result of each classification agent into a recommendation agent, generating a dialogue through the recommendation agent and recommending the generated dialogue to the user, so as to realize the dialogue recommendation to the user through cooperation between agents in the multi-agent system.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a recommended processing method and apparatus. Background Technology

[0002] With the continuous development and promotion of the Internet and artificial intelligence, many services can be automated through online deployed intelligent agents. For example, when users access online services, the online services can make targeted recommendations to users based on the deployed intelligent agents. However, as more and more online services make recommendations based on various forms of intelligent agents, the competition among all parties is becoming more and more intense. In this case, higher requirements are placed on access recommendations based on intelligent agents. Summary of the Invention

[0003] This specification provides one or more embodiments of a recommendation processing method applied to a multi-agent system. The method includes: acquiring dialogue data of users interacting in a resource service; classifying and recommending the dialogue data by scheduling agents, and sending the dialogue data to each classification agent after a successful match; having each classification agent perform task detection on the dialogue data for its respective category recommendation task to obtain detection results; inputting the obtained detection results into the recommendation agent to generate a dialogue and recommending it to the user.

[0004] This specification provides one or more embodiments of a recommendation processing apparatus operating in a multi-agent system. The apparatus includes: a dialogue acquisition module configured to acquire dialogue data of users interacting in a resource service; a recommendation matching module configured to perform classification recommendation matching on the dialogue data through scheduling agents, and send the dialogue data to each classification agent after successful matching; a classification detection module configured to perform task detection on the dialogue data for each classification recommendation task through the classification agents to obtain detection results; and a dialogue generation module configured to input the obtained detection results into the recommendation agents to generate dialogue and recommend dialogue to the users.

[0005] This specification provides one or more embodiments of a recommendation processing device, including: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to: acquire dialogue data of users interacting in a resource service; perform classification and recommendation matching on the dialogue data by scheduling agents, and send the dialogue data to each classification agent after a successful match; perform task detection on the dialogue data by each classification agent to obtain detection results for each category recommendation task; input the obtained detection results into the recommendation agent to generate dialogue and recommend dialogue to the user.

[0006] This specification provides one or more embodiments of a computer-readable storage medium for storing computer-executable instructions, which, when executed, perform the following process: acquiring dialogue data of users interacting in a resource service; classifying and recommending the dialogue data through a scheduling agent, and sending the dialogue data to each classification agent after a successful match; performing task detection on the dialogue data for each classification recommendation task through each classification agent to obtain detection results; inputting the obtained detection results into a recommendation agent to generate a dialogue and recommending a dialogue to the user. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in one or more embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram illustrating the implementation environment of a recommended processing method provided in one or more embodiments of this specification;

[0009] Figure 2 A flowchart illustrating a recommended processing method provided in one or more embodiments of this specification;

[0010] Figure 3 A flowchart illustrating a recommendation processing method applied to a resource service scenario, provided for one or more embodiments of this specification;

[0011] Figure 4 A schematic diagram of a recommended processing apparatus embodiment provided by one or more embodiments of this specification;

[0012] Figure 5 This is a schematic diagram of the structure of a recommended processing device provided for one or more embodiments of this specification. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0014] The recommended processing methods provided in one or more embodiments of this specification are applicable to the implementation environment of a recommended processing system. (Refer to...) Figure 1 The implementation environment includes at least:

[0015] A multi-agent system 100, comprising: a scheduling agent 101-1, a classification agent 102, and a recommendation agent 103; furthermore, the implementation environment may also include a large language model 200, which is used to respond to calls from each agent in the multi-agent system 100 and perform corresponding processing.

[0016] Specifically, the scheduling agent 101-1 is used to perform preliminary classification and recommendation matching of the user's dialogue data, and further distribute the matched dialogue data to the classification agent 102;

[0017] The number of classification agents 102 can be one or more. That is, classification agents 102 can be composed of classification agents 102-1, or they can be composed of classification agents 102-1 to classification agents 102-N. Each classification agent is used to perform task detection of its own classification recommendation task on the user's dialogue data and obtain the detection results of each classification agent.

[0018] The recommended agent 103 is used to generate dialogues and recommend dialogues to users based on the detection results of each class of agents.

[0019] In this implementation environment, during the process of recommending dialogues to users when they access resource services, the system first acquires dialogue data from the user's interactions during resource service access. Then, it uses a scheduling agent 101-1 to classify and match the user's dialogue data. This classification matching determines whether to use a classification agent to recommend the user's dialogue data accordingly. Matched dialogue data is then sent to each classification agent. Each classification agent, upon receiving the dialogue data, performs task detection for the classification recommendation task to obtain its own detection results. These results are then input into a recommendation agent 103. The recommendation agent 103 generates dialogues based on the input detection results and recommends the generated dialogues to the user. This multi-agent system achieves dialogue recommendation to the user through cooperation between agents.

[0020] It should be noted that, considering that the user's conversation data, service access data, user data, and other related data involved in this specification may, to some extent, constitute user privacy, authorization from the user must be obtained before collecting such data to ensure that the data collection operation complies with relevant data management regulations. For example, data authorization can be granted when the user first accesses the resource service, during any subsequent access to the resource service, or when the user triggers a service identifier. Specific methods of data authorization include sending a user data authorization reminder to the user, who then confirms the reminder with an instruction, or obtaining user data authorization by signing a data authorization agreement.

[0021] One or more embodiments of a recommended processing method provided in this specification are as follows:

[0022] Reference Figure 2 The recommended processing method provided in this embodiment can be applied to multi-agent systems. The method specifically includes steps S202 to S208.

[0023] Step S202: Obtain the dialogue data of users who are interacting with each other in the resource service.

[0024] The resource service mentioned in this embodiment refers to resource-related services provided within the application, such as resource services related to insurance projects provided by the application. Users can access these resource services to consult, inquire about, purchase, and / or sell purchased insurance projects. They can also apply for insurance coverage or perform other insurance-related processing through these resource services. Insurance projects can be resource management insurance projects that provide protection for users' resource management in the application, such as financial insurance projects.

[0025] The dialogue data refers to the data generated by users during dialogue interactions when accessing resource services. Specifically, the dialogue data can be the data generated by users engaging in dialogue within the interaction area following the triggering of the resource service's service identifier.

[0026] Optionally, dialogue interaction is initiated by triggering the service identifier of the resource service. The service identifier is a identifier configured by the resource service for interacting with the user. Once triggered, an interaction area is displayed where the user can inquire, request prices, and / or make purchases. Additionally, the resource service can also recommend resources or engage in dialogue with the user within the interaction area. Optionally, the service identifier may include the interaction identifier of a virtual or digital human configured by the resource service.

[0027] For example, the dialogue data generated when a user interacts with the digital human configured in the insurance project service is as follows:

[0028] Digital Human: Have you ever purchased any health insurance programs before?

[0029] User: No

[0030] User: I have access to health insurance programs from other organizations.

[0031] Digital Human: Do you have any recent health insurance claims?

[0032] User: No

[0033] Digital Human: How much health insurance coverage would you like to purchase?

[0034] Users: 300,000.

[0035] In practice, during the process of a user accessing a resource service and engaging in dialogue interaction, the system acquires dialogue data obtained from the user's dialogue interaction within the resource service. Based on this dialogue data, the system recommends dialogues to the user during the user's access to the resource service.

[0036] Step S204: The dialogue data is classified and recommended for matching by scheduling agents, and the dialogue data is sent to each classification agent after the matching is successful.

[0037] The scheduling agent mentioned in this embodiment refers to the agent responsible for scheduling in a multi-agent system; the classification agent refers to the agent in the multi-agent system that performs recommendation detection processing for each category; wherein, after the scheduling agent performs classification recommendation matching, the obtained matching results include successful matching and unsuccessful matching. Successful matching indicates that further classification recommendation processing is needed for the user's dialogue data. For successfully matched dialogue data, the successfully matched dialogue data is sent to each classification agent so that the corresponding category recommendation detection processing can be performed in each classification agent;

[0038] A failed match indicates that no further classification and recommendation processing is needed for the user's dialogue data. For dialogue data that fails to match, you can either respond to the dialogue data or leave it unprocessed. For example, responding to the dialogue data could involve inputting the dialogue data into a dialogue generation model to generate a dialogue and then recommending the generated dialogue to the user.

[0039] In practice, after obtaining the user's dialogue data, the system schedules intelligent agents to classify and recommend the user's dialogue data to determine whether the current dialogue data needs further classification and recommendation processing. If the match is successful, the dialogue data is sent to each classification intelligent agent, that is, the matched dialogue data is sent to each classification intelligent agent so that each classification intelligent agent can perform further classification and recommendation processing on the matched dialogue data.

[0040] Specifically, in one optional implementation of this embodiment, classifying and recommending user dialogue data includes:

[0041] Perform intent recognition on user dialogue data to obtain dialogue intent, or perform semantic recognition on user dialogue data to obtain dialogue semantics;

[0042] Detect the matching degree between the dialogue intent or dialogue semantics and the classification recommendation task of each classification agent. If any matching degree is greater than a preset threshold, the matching is determined to be successful.

[0043] Furthermore, if the overall matching degree is less than or equal to the preset threshold, it indicates that no further classification and recommendation processing of the user's dialogue data is required. In this case, dialogue response to the dialogue data can be performed or no processing can be performed. For example, dialogue response to the dialogue data can be performed by inputting the dialogue data into the dialogue generation model to generate dialogue and recommending the generated dialogue to the user.

[0044] It should be noted that the process of classifying and recommending matching user dialogue data can be achieved through a large language model. For example, a large language model can be used to classify and recommend matching user dialogue data to obtain matching results, which include whether the match is successful or unsuccessful. Specifically, matching text containing dialogue data and classification and recommendation matching fields can be created. The matching text containing dialogue data and classification and recommendation matching fields can be input into the large language model to perform classification and recommendation matching and output matching results. In addition, the large language model can be fine-tuned or trained in advance so that it has the ability to perform classification and recommendation matching and output matching results. In this way, dialogue data can be input into the large language model to perform classification and recommendation matching and output matching results.

[0045] Similarly, the above-mentioned process of obtaining dialogue intent by performing intent recognition on user dialogue data and detecting the matching degree between the dialogue intent and the classification recommendation task of each classification agent can also be implemented using a large language model. For example, a text containing dialogue data and processing fields for performing dialogue intent recognition and matching degree calculation can be created. The intent recognition is performed by inputting the text containing dialogue data and processing fields into the large language model, and the matching degree between the dialogue intent and the classification recommendation task of each classification agent is calculated and the matching degree is output.

[0046] Alternatively, the above-mentioned method of performing semantic recognition on user dialogue data to obtain dialogue semantics and detecting the matching degree between dialogue semantics and the classification recommendation tasks of each classification agent can also be implemented using a large language model. For example, a text containing dialogue data and processing fields for dialogue semantic recognition and matching degree calculation can be created. Semantic recognition can be performed by inputting the text containing dialogue data and processing fields into a large language model, and the matching degree between dialogue semantics and the classification recommendation tasks of each classification agent can be calculated and the matching degree can be output.

[0047] Step S206: The various classification agents perform task detection on the dialogue data for each category recommendation task to obtain detection results.

[0048] In this embodiment, multiple classification agents are set up in the multi-agent system. Each classification agent is used to perform recommendation detection processing on the user's dialogue data in the corresponding category. Specifically, the recommendation processing performed by each classification agent includes task detection of its own category recommendation task on the user's dialogue data. That is, each classification agent realizes the recommendation detection processing of the user's dialogue data by performing task detection of its own category recommendation task on the user's dialogue data.

[0049] Optionally, there is a one-to-one correspondence between classification agents and classification recommendation tasks, with each classification agent performing task execution detection for its corresponding classification recommendation task.

[0050] Among them, the classification recommendation task refers to the dialogue recommendation task in which a classification agent makes dialogue recommendations based on the user's dialogue data under a single classification (dimension). The task detection of the classification recommendation task refers to detecting whether to make dialogue recommendations based on the user's dialogue data. Here, the task detection of the classification recommendation task is a binary classification detection of the classification recommendation task. That is, there are two detection results for the classification recommendation task: one is to make a recommendation, and the other is not to make a recommendation.

[0051] Optional, categorized recommendation tasks include: a payment change recommendation task that recommends to users to increase the payment period and payment method of the coverage items, and / or an increase recommendation task that recommends to users to increase the coverage amount of the coverage items.

[0052] For example, when a user accesses a resource service guarantee project, during the process of the user clicking the service icon of the resource service and interacting with the digital human of the resource service, the digital human will determine whether to make a dialogue recommendation to the user based on the dialogue data generated by the user and the digital human. Here, the classification recommendation task corresponding to the classification agent A in the multi-agent system is: to recommend to the user to increase the payment period of the guarantee project and change the payment method. Specifically, this classification recommendation task is to recommend to the user to change the monthly payment method of the guarantee project to a quarterly payment method or an annual payment method.

[0053] In the multi-agent system, the classification recommendation task corresponding to the classification agent B is: to recommend to the user to increase the coverage amount of the insurance project. Specifically, this classification recommendation task is to recommend to the user to increase the coverage amount of the insurance project currently selected by the user (e.g., 100,000) to a preset coverage amount (e.g., 300,000).

[0054] In practice, each category agent performs task detection on the user's dialogue data for each category recommendation task. Specifically, each category agent performs task detection on the user's dialogue data for the category recommendation task corresponding to that category agent, and obtains the detection results of each category agent.

[0055] Specifically, in the process of performing task detection for each category recommendation task on user dialogue data, task detection for each category recommendation task can also be performed using a large language model. In one optional implementation method provided in this embodiment, the task detection results for each category recommendation task on user dialogue data are obtained, including:

[0056] Generate detection text for each task based on the dialogue data and the task detection templates for each category of recommended tasks;

[0057] The detection texts for each task are input into a large language model to perform task execution detection for the classification and recommendation tasks of the task detection text records, and the detection results are obtained.

[0058] The task detection template for the classification recommendation task refers to the descriptive text used to instruct the large language model to perform task detection for the classification recommendation task. For example, in the task detection template for the classification recommendation task corresponding to the classification agent A in a multi-agent system, the descriptive statement is: "You are an AI assistant familiar with insurance sales. Please determine whether it is the right time to recommend the quarterly or annual payment method of the insurance project to the user after the user's last sentence of dialogue, as it is more advantageous than the monthly payment method?"

[0059] For example, in the task detection template for the classification recommendation task of classification agent B in a multi-agent system, the description statement is: "You are an AI assistant familiar with insurance sales. Please determine whether it is the right time to suggest that the user increase the coverage amount of the currently selected insurance program to the preset coverage amount after the user's last sentence?"

[0060] In practical applications, to improve the effectiveness and accuracy of dialogue recommendations based on user dialogue data, and to increase the conversion rate of these recommendations, execution condition statements for each category of recommendation task can be set in the task detection template. This means that by setting execution condition statements for task detection in the task detection template, the large language model is prompted to perform more effective and accurate task detection based on these condition statements. This allows the timing of dialogue recommendations to better match user needs, thus helping to improve the conversion rate of dialogue recommendations.

[0061] Optionally, the task detection template for each category recommendation task records at least one execution condition statement for each category recommendation task; that is, the task detection template for each category recommendation task records at least one execution condition statement for that category recommendation task.

[0062] For example, in a multi-agent system, the execution condition statements recorded in the task detection template for the classification recommendation task corresponding to classification agent A include:

[0063] "1. This opportunity can be triggered when the user's last sentence is asking about the payment method, such as annual payment or monthly payment;

[0064] 2. This opportunity can be triggered when the user's last sentence indicates their intention to pay monthly.

[0065] 3. The annual payment option has already been recommended to the user during previous conversations, so this opportunity should not be triggered.

[0066] 4. If the user has already inquired about payment methods during the conversation, this opportunity cannot be triggered.

[0067] 5. If it can be clearly determined during the conversation that the current coverage product is not the designated health-related coverage product, then this event cannot be triggered.

[0068] Specifically, based on the task detection template of each classification recommendation task recording at least one execution condition statement for that classification recommendation task, the large language model can perform binary classification detection on each execution condition statement during the task execution detection process of the classification recommendation task recorded in the task detection text. Finally, the detection of the classification recommendation task is determined based on the detection results of each binary classification detection.

[0069] In one optional implementation of this embodiment, task execution detection for classifying and recommending text records for task detection includes:

[0070] Read the description statement and at least one execution condition statement of the classification recommendation task from the classification detection text records;

[0071] The system performs binary classification detection on whether the dialogue data triggers the description statement and each execution condition statement, and determines the detection result based on the binary classification detection result.

[0072] In this embodiment, under the implementation of task detection for each category recommendation task using a large language model, before the large language model performs task detection for each category recommendation task, a large language model capable of performing task detection for each category recommendation task can be obtained by fine-tuning the pre-trained model. Specifically, in an optional implementation provided in this embodiment, the large language model is trained in the following way: a training text containing task records of each category recommendation agent is created, and the training text is input into the pre-trained model. The pre-trained model is then fine-tuned based on the training text to obtain the large language model.

[0073] A pre-trained model refers to a natural language model that has been trained in advance. Pre-trained models can adopt neural network architectures with a large number of parameters. Pre-trained models can adopt pre-trained large language models (LLMs), such as chatGPT (chat Generative Pre-trained Transformer) and other open-source large language models.

[0074] It should be noted that the classification recommendation task can be a binary classification recommendation task, meaning that the task detection result of each classification recommendation task is either to recommend or not to recommend. Multiple binary classification recommendation tasks can be obtained by splitting multi-label classification recommendation tasks, or vice versa. Here, a multi-label classification recommendation task refers to a dialogue recommendation task that classifies user dialogue data under multiple categories (dimensions) and finally determines one or more category labels for dialogue recommendation; a multi-class recommendation task refers to a dialogue recommendation task that classifies user dialogue data under multiple categories and finally determines one category for dialogue recommendation. Based on this, we will provide users with... When a multi-label classification recommendation task or a multi-class recommendation task is broken down into multiple binary classification recommendation tasks, the probability of each binary classification recommendation task being triggered is relatively low. Therefore, if a pre-trained model is fine-tuned to obtain a large language model using a multi-label classification recommendation task or a multi-class recommendation task approach, a large number of training samples are required for training each category. However, if a pre-trained model is fine-tuned to obtain a large language model using a binary classification recommendation task approach, the number of training samples required is greatly reduced. Only corresponding training samples need to be constructed specifically for each binary classification recommendation task.

[0075] Step S208: Input the obtained detection results into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0076] The above-mentioned task detection is performed on the user's dialogue data by each category intelligent agent to obtain the detection results of each category recommendation task, that is, the detection results of each category intelligent agent. Based on this, the obtained detection results of each category recommendation task are input into the recommendation intelligent agent, which generates dialogue and recommends dialogue to the user.

[0077] As mentioned above, the detection results of each classifying agent include either making a recommendation or not making a recommendation. Among the multiple detection results of multiple classifying agents, there may be one or more classifying agents whose detection results are making a recommendation, or there may be a situation where all classifying agents' detection results are not making a recommendation. In this case, the dialogue data can be responded to in a dialogue manner or no processing can be performed. For example, responding to the dialogue data in a dialogue manner can be done by inputting the dialogue data into the dialogue generation model to generate a dialogue and then recommending the generated dialogue to the user.

[0078] Specifically, in the process of recommending dialogues to users by a recommendation agent, when only one of the multiple detection results from multiple classification agents indicates that a recommendation should be made, this embodiment provides an optional implementation method in which dialogue generation and recommendation to users includes:

[0079] If each detection result contains only one detection result that is recommended, generate a recommendation dialogue for the category recommendation task corresponding to that detection result that is recommended, and determine the recommendation node of the recommendation dialogue so as to send the recommendation dialogue to the user according to the recommendation node;

[0080] or,

[0081] If each detection result contains only one recommended detection result, a recommendation dialogue for the category recommendation task corresponding to that recommended detection result is generated, and the merged recommendation dialogue is displayed in the dialogue area triggered by the service identifier of the resource service.

[0082] Furthermore, in cases where at least two of the detection results from multiple classification agents indicate a recommendation, this embodiment provides an optional implementation method that generates a dialogue and provides a dialogue recommendation to the user, including:

[0083] If each detection result contains at least two detection results that are recommended, a merged recommendation dialogue is generated based on the dialogue data and at least two category recommendation tasks corresponding to the at least two detection results that are recommended, and the merged recommendation dialogue is displayed in the dialogue area after the service identifier of the resource service is triggered.

[0084] or,

[0085] If each detection result contains at least two detection results that are recommended, a merged recommendation dialogue is generated based on the dialogue data and at least two category recommendation tasks corresponding to the at least two detection results that are recommended. The recommendation node of the merged recommendation dialogue is determined so as to send the merged recommendation dialogue to the user according to the recommendation node.

[0086] Optionally, the merged recommended dialogue generated by the recommended agent or the recommended dialogue can be displayed in the interactive area after the service identifier is triggered.

[0087] In the specific execution process, the recommendation dialogue for the classification recommendation task corresponding to the detection results for making recommendations can be generated through a dialogue generation model. The dialogue generation model can adopt a large language model. Specifically, the dialogue generation text containing the classification recommendation task corresponding to the detection results and the dialogue generation statement can be input into the dialogue generation model to generate the recommendation dialogue. Similarly, the above-mentioned generation of a merged recommendation dialogue based on the dialogue data and at least two classification recommendation tasks corresponding to at least two detection results for making recommendations can be generated through a dialogue generation model. Specifically, the dialogue generation text containing at least two classification recommendation tasks and the merged dialogue generation statement can be input into the dialogue generation model to generate the merged recommendation dialogue.

[0088] It should be noted that the above implementation provides multiple machine learning models, specifically including: a large language model for task execution detection based on input task detection text, a dialogue generation model for dialogue generation based on input dialogue data, and a dialogue generation model for dialogue generation based on input dialogue generation text. The dialogue generation model for dialogue generation based on input dialogue data and the dialogue generation model for dialogue generation based on input dialogue generation text can be the same dialogue generation model, and this dialogue generation model can be a large language model. In this case, the large language model for dialogue generation and the large language model for task execution detection can be the same large language model or different large language models.

[0089] In one or more embodiments, the display of the service identifier may be determined based on the user's service access data and / or user data, or based on other data, such as service data of resource services and / or resource data of resources. Alternatively, the display of the service identifier may be determined based on at least one of service access data and user data and at least one of service data and resource data of resources.

[0090] Specifically, the display of service identifiers can correspond to resource services or service pages of resource services. For example, service identifiers can be displayed on specific service pages of resource services, but not on other service pages of resource services. Specifically, the specific service page for displaying service identifiers can be determined by inputting service access data, user data, service data of resource services and / or resource data of resources into a pre-trained identifier decision model or a large language model.

[0091] In addition, the display of service identifiers can also correspond to resources within resource services. For example, the service identifier may be displayed on the resource page of a specific resource within a resource service, while it may not be displayed on the resource pages of other resources. Specifically, this can be achieved by inputting user service access data, user data, resource service service data, and / or resource data into a pre-trained identifier decision model or large language model to determine the specific resource for which the service identifier is displayed; or by inputting service access data, user data, resource service service data, and / or resource data into a pre-trained identifier decision model or large language model to determine the resource page for which the service identifier is displayed; or, the service... The display of identifiers can also correspond to users. For example, a service identifier can be displayed during a specific user's access process. Specifically, the user's service access data, user data, service data of resource services, and / or resource data of resources can be input into a pre-trained identifier decision model or a large language model to determine whether to display a service identifier for that user. Alternatively, the display of service identifiers can also correspond to at least two of the following: resource services, service pages of resource services, resources, and users. Similarly, the display of service identifiers can be determined by inputting the user's service access data, user data, service data of resource services, and / or resource data of resources into a pre-trained identifier decision model or a large language model.

[0092] In summary, the recommendation processing method provided in this embodiment, based on a user's access to a resource service and dialogue interaction, obtains the dialogue data obtained by the user during the dialogue interaction on the resource service. It then uses a scheduling agent to classify and match the user's dialogue data to determine whether further classification and recommendation processing is needed. If the match is successful, the dialogue data is sent to each classification agent. Upon receiving the sent dialogue data, each classification agent performs task detection for its corresponding classification recommendation task on the user's dialogue data, obtaining the detection results of each classification agent. Furthermore, the obtained detection results of each classification recommendation task are input into the recommendation agent, which then generates dialogue and recommends it to the user. This multi-agent system, through the cooperation of multiple agents, improves the effectiveness and accuracy of dialogue recommendations to the user, allowing the timing of dialogue recommendations to better match the user's needs and helping to increase the conversion rate of dialogue recommendations.

[0093] The following example uses a recommendation processing method provided in this embodiment in a resource service scenario as an example, combined with... Figure 3 The recommended processing method provided in this embodiment will be further explained below. Figure 3The recommendation processing method applied to resource service scenarios includes the following steps.

[0094] Step S302: Obtain the dialogue data of users who are interacting with each other in the resource service.

[0095] Step S304: The scheduling agent in the multi-agent system classifies, recommends, and matches the dialogue data to obtain matching results.

[0096] Step S306: If the matching result is a successful match, send dialogue data to the first and second classifying agents in the multi-agent system.

[0097] Step S308: The first classification agent obtains the first detection result by performing a first classification recommendation task on the dialogue data.

[0098] Step S310: The second classification agent obtains the second detection result by performing a second classification recommendation task on the dialogue data.

[0099] Step S312: Input the first detection result and the second detection result into the recommended agent in the multi-agent system.

[0100] Step S314: If the first detection result or the second detection result is to make a recommendation, the recommendation agent generates the corresponding recommendation dialogue for the first category recommendation task or the second category recommendation task and sends it to the user.

[0101] Step S316: If both the first and second detection results are to make a recommendation, the recommendation agent generates a combined recommendation dialogue of the first and second classification recommendation tasks and sends it to the user.

[0102] It should be noted that any one or more steps in steps S302 to S316 can be combined with any one or more steps in steps S202 to S208 to form a new implementation method according to the needs of implementation and deployment. In addition, any one or more technical features in steps S302 to S316 can be selected and combined with any one or more technical features provided in steps S202 to S208 to form a new implementation method according to the actual deployment needs. Alternatively, any one or more technical features in steps S302 to S316 can be replaced with any one or more technical features provided in steps S202 to S208 to form a new implementation method according to the actual deployment needs. These will not be elaborated on here.

[0103] The following is an embodiment of a recommended processing device provided in this specification:

[0104] In the above embodiments, a recommendation processing method is provided, and correspondingly, a recommendation processing apparatus is also provided, which will be described below with reference to the accompanying drawings.

[0105] Reference Figure 4 This illustration shows a schematic diagram of a recommended processing apparatus embodiment provided in this embodiment.

[0106] Since the apparatus embodiments correspond to the method embodiments, the descriptions are relatively simple. For relevant parts, please refer to the corresponding descriptions of the method embodiments provided above. The apparatus embodiments described below are merely illustrative.

[0107] This embodiment provides a recommendation processing device, the device comprising:

[0108] The dialogue acquisition module 402 is configured to acquire dialogue data of users who are engaging in dialogue interactions in the resource service;

[0109] The recommendation matching module 404 is configured to classify and recommend matching the dialogue data through a scheduling agent, and send the dialogue data to each classification agent after a successful match.

[0110] The classification detection module 406 is configured to perform task detection of each category recommendation task on the dialogue data by the respective classification agents to obtain detection results.

[0111] The dialogue generation module 408 is configured to input the obtained detection results into the recommendation agent to generate dialogue and recommend dialogue to the user.

[0112] The following is an example of a recommended processing device provided in this specification:

[0113] Corresponding to the recommendation processing method described above, based on the same technical concept, one or more embodiments of this specification also provide a recommendation processing device for executing the recommendation processing method provided above. Figure 5 This is a schematic diagram of the structure of a recommended processing device provided for one or more embodiments of this specification.

[0114] This embodiment provides a recommended processing device, including:

[0115] like Figure 5As shown, the recommended processing device can vary considerably due to differences in configuration or performance. It may include one or more processors 501 and memory 502, where one or more application programs or data may be stored. The memory 502 may be temporary or persistent storage. The application programs stored in the memory 502 may include one or more modules (not shown), each module including a series of computer-executable instructions from the recommended processing device. Furthermore, the processor 501 may be configured to communicate with the memory 502 and execute the series of computer-executable instructions stored in the memory 502 on the recommended processing device. The recommended processing device may also include one or more power supplies 503, one or more wired or wireless network interfaces 504, one or more input / output interfaces 505, one or more keyboards 506, etc.

[0116] In one specific embodiment, the recommended processing device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the recommended processing device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0117] Acquire dialogue data of users who engage in conversational interactions within the resource service;

[0118] The dialogue data is classified, recommended, and matched by scheduling agents, and the dialogue data is sent to each classification agent after a successful match.

[0119] The detection results are obtained by performing task detection for each category recommendation task on the dialogue data by the aforementioned intelligent agents.

[0120] The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0121] This specification provides an embodiment of a computer-readable storage medium as follows:

[0122] Corresponding to the recommended processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a computer-readable storage medium.

[0123] The computer-readable storage medium provided in this embodiment is used to store computer-executable instructions, which, when executed, implement the following process:

[0124] Acquire dialogue data of users who engage in conversational interactions within the resource service;

[0125] The dialogue data is classified, recommended, and matched by scheduling agents, and the dialogue data is sent to each classification agent after a successful match.

[0126] The detection results are obtained by performing task detection for each category recommendation task on the dialogue data by the aforementioned intelligent agents.

[0127] The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0128] It should be noted that the embodiments of a computer-readable storage medium described in this specification and the embodiments of a recommended processing method described in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding method described above, and the repeated parts will not be described again.

[0129] This specification provides an example of a computer program product as follows:

[0130] Corresponding to the recommended processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a computer program product.

[0131] A computer program product includes a computer program / instructions that, when executed by a processor, perform the following steps:

[0132] Acquire dialogue data of users who engage in conversational interactions within the resource service;

[0133] The dialogue data is classified, recommended, and matched by scheduling agents, and the dialogue data is sent to each classification agent after a successful match.

[0134] The detection results are obtained by performing task detection for each category recommendation task on the dialogue data by the aforementioned intelligent agents.

[0135] The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0136] It should be noted that the embodiments of a computer program product described in this specification and the embodiments of a recommended processing method described in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding method described above, and the repeated parts will not be described again.

[0137] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments. For example, the device embodiments, equipment embodiments, computer-readable storage medium embodiments, and computer program product embodiments are all similar to the method embodiments, so the descriptions are relatively simple. For reading the relevant content of the device embodiments, equipment embodiments, computer-readable storage medium embodiments, and computer program product embodiments, please refer to the description of the method embodiments.

[0138] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0139] In the 1930s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement to the methodology cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0140] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0141] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0142] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0143] Those skilled in the art will understand that one or more embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0145] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] These computer program instructions may also be loaded onto a computer or other programmable processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0147] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0148] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0149] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0150] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising at least one…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0151] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0152] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A recommendation processing method applied to a multi-agent system, the method comprising: Acquire dialogue data of users who engage in conversational interactions within the resource service; The dialogue data is classified, recommended, and matched by scheduling agents, and the dialogue data is sent to each classification agent after a successful match. The following operations are performed by each classifying agent: generating task detection text based on the dialogue data and the task detection template of each classifying recommendation task and inputting it into the large language model; reading the description statement and at least one execution condition statement of the classifying recommendation task recorded in the classification detection text; performing binary classification detection on whether the dialogue data triggers the description statement and each execution condition statement and determining the detection result. The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

2. The recommendation processing method according to claim 1, wherein each classification agent corresponds one-to-one with the classification recommendation task, and each classification agent performs task execution detection for its corresponding classification recommendation task; The task execution detection includes binary classification detection to determine whether the dialogue data triggers the task execution conditions for the categorized recommendation task.

3. The recommendation processing method according to claim 2, wherein the classification recommendation task includes: The task recommends a payment change to the user by increasing the payment period and payment method of the coverage items, and / or recommends an increase in the coverage amount of the coverage items to the user.

4. The recommendation processing method according to claim 1, wherein the large language model is trained in the following manner: Training text is created that contains task records of the classification recommendation tasks of each classification agent, and the training text is input into a pre-trained model. The pre-trained model is then fine-tuned based on the training text to obtain the large language model.

5. The recommendation processing method according to claim 1, wherein generating a dialogue and recommending a dialogue to the user comprises: If only one of the detection results is recommended, a recommendation dialogue for the category recommendation task corresponding to the recommended detection result is generated, and the recommendation node of the recommendation dialogue is determined so as to send the recommendation dialogue to the user according to the recommendation node.

6. The recommendation processing method according to claim 1, wherein generating a dialogue and recommending a dialogue to the user comprises: If each detection result contains at least two recommended detection results, a merged recommendation dialogue is generated based on the dialogue data and the at least two category recommendation tasks corresponding to the at least two recommended detection results, and the merged recommendation dialogue is displayed in the dialogue area after the service identifier of the resource service is triggered.

7. The recommendation processing method according to claim 1, wherein classifying and recommending the dialogue data includes: The dialogue data can be subjected to intent recognition to obtain the dialogue intent, or the dialogue data can be subjected to semantic recognition to obtain the dialogue semantics. The matching degree between the dialogue intent or the dialogue semantics and the classification recommendation task of each classification agent is detected. If any matching degree is greater than a preset threshold, the matching is determined to be successful.

8. The recommendation processing method according to claim 7, after the step of performing the classification, recommendation, and matching operation on the dialogue data by scheduling the intelligent agent, it further includes: If the matching degree is less than or equal to the preset threshold, the dialogue data is input into the dialogue generation model to generate a dialogue, and the generated dialogue is recommended to the user.

9. The recommendation processing method according to claim 1, wherein the dialogue interaction is performed by triggering the service identifier of the resource service; The dialogue generated by the recommending agent is displayed in the interactive area after the service identifier is triggered.

10. A recommendation processing apparatus, operating in a multi-agent system, the apparatus comprising: The dialogue acquisition module is configured to acquire dialogue data of users who engage in dialogue interactions within the resource service; The recommendation matching module is configured to classify and recommend matching the dialogue data by scheduling agents, and send the dialogue data to each classification agent after a successful match. The classification detection module is configured to perform the following operations through each classification agent: generate task detection texts based on the dialogue data and the task detection templates of each category recommendation task and input them into the large language model; read the description statement and at least one execution condition statement of the category recommendation task recorded in the classification detection text; perform binary classification detection on whether the dialogue data triggers the description statement and each execution condition statement and determine the detection result. The dialogue generation module is configured to input the obtained detection results into the recommendation agent to generate dialogue and recommend dialogue to the user.

11. A recommended processing apparatus, comprising: processor; And, a memory configured to store computer-executable instructions, which, when executed, cause the processor to: Acquire dialogue data of users who engage in conversational interactions within the resource service; The dialogue data is classified, recommended, and matched by scheduling agents, and the dialogue data is sent to each classification agent after a successful match. The following operations are performed by each classifying agent: generating task detection text based on the dialogue data and the task detection template of each classifying recommendation task and inputting it into the large language model; reading the description statement and at least one execution condition statement of the classifying recommendation task recorded in the classification detection text; performing binary classification detection on whether the dialogue data triggers the description statement and each execution condition statement and determining the detection result. The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

12. A computer-readable storage medium for storing computer-executable instructions that, when executed, implement the steps of the method of claim 1.

Images