Recommendation processing method and device

Abstract

The embodiment of the invention provides a recommendation processing method and device.The recommendation processing method comprises the steps that in the process of conducting dialogue recommendation on a user through a multi-agent system when the user accesses a resource service, classified recommendation matching is conducted on dialogue data of the user through a scheduling agent, and the matching result is obtained; and sending the dialogue data passing matching to each classification agent, performing task detection of a classification recommendation task on the dialogue data through each classification agent to obtain a detection result, inputting the detection result of each classification agent into a recommendation agent, performing dialogue generation through the recommendation agent, and recommending the generated dialogue to a user. Therefore, dialogue recommendation to the user is realized through cooperation among the intelligent agents in the multi-intelligent-agent system.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a recommendation processing method and device. Background Art

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

[0003] One or more embodiments of the present specification provide a recommendation processing method, which is applied to a multi-agent system, and the method includes: obtaining conversation data of users who are interacting with each other in a resource service. Classifying and recommending the conversation data through a scheduling agent, and sending the conversation data to each classified agent after the match is passed. Performing task detection of each classified recommendation task on the conversation data by each classified agent to obtain a detection result. Inputting each detection result obtained into a recommendation agent to generate a conversation and make a conversation recommendation to the user.

[0004] One or more embodiments of the present specification provide a recommendation processing device that runs on a multi-agent system, the device comprising: a conversation acquisition module, configured to acquire conversation data of users who are interacting in conversations with resource services. A recommendation matching module, configured to classify and recommend the conversation data by scheduling an agent, and send the conversation data to each classified agent after the match is successful. A classification detection module, configured to obtain a detection result by performing task detection of each classified recommendation task on the conversation data by each classified agent. A conversation generation module, configured to input each detection result obtained into a recommendation agent for conversation generation and make conversation recommendations to the user.

[0005] One or more embodiments of the present specification provide a recommendation processing device, including: a processor; and a memory configured to store computer executable instructions, wherein when the computer executable instructions are executed, the processor: obtains the conversation data of users who are interacting with each other in a resource service. The conversation data is classified and recommended by a scheduling agent, and the conversation data is sent to each classification agent after the match is passed. The task detection of each classification recommendation task on the conversation data is performed by each classification agent to obtain a detection result. The detection results are input into the recommendation agent to generate a conversation and make a conversation recommendation to the user.

[0006] One or more embodiments of the present specification provide a computer-readable storage medium for storing computer-executable instructions, which implement the following process when executed: obtaining conversation data of users who are interacting with each other in a resource service. Classifying and recommending the conversation data through a scheduling agent, and sending the conversation data to each classification agent after the match is passed. Performing task detection of each classification recommendation task on the conversation data by each classification agent to obtain a detection result. Inputting each detection result obtained into a recommendation agent to generate a conversation and make a conversation recommendation to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate one or more embodiments of the present specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present specification. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative labor. Figure 1 A schematic diagram of an implementation environment of a recommended processing method provided for one or more embodiments of this specification; Figure 2 A processing flow chart of a recommended processing method provided for one or more embodiments of this specification; Figure 3 A processing flow chart of a recommended processing method applied to a resource service scenario provided in one or more embodiments of this specification; Figure 4 A schematic diagram of a recommended processing device embodiment provided for one or more embodiments of this specification; Figure 5 A schematic diagram of the structure of a recommended processing device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0008] In order 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 in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0009] The recommendation processing method provided in one or more embodiments of this specification can be applied to the implementation environment of the recommendation processing system. Figure 1 , the implementation environment at least includes: A multi-agent system 100, wherein the multi-agent system 100 includes: a scheduling agent 101-1, a classification agent 102 and a recommendation agent 103; in addition, the implementation environment may also include a large language model 200, which is used to respond to the call of each agent in the multi-agent system 100 and perform corresponding processing; Specifically, the scheduling agent 101-1 is used to perform preliminary classification recommendation matching on the user's conversation data, and further distribute the matching conversation data to the classification agent 102; The number of the classification agents 102 may be one or more, that is, the classification agents 102 may be composed of the classification agent 102-1, and may also be composed of the classification agents 102-1 to 102-N, each of which is used to perform task detection of its own classification recommendation task on the user's conversation data to obtain the detection results of each classification agent; The recommendation agent 103 is used to generate dialogues based on the detection results of each classification agent and make dialogue recommendations to the user.

[0010] In this implementation environment, when making conversation recommendations to users when they access resource services, conversation data of users' conversation interactions during the process of accessing resource services is first obtained, and the user's conversation data is classified and recommended by the scheduling agent 101-1. The classification recommendation matching is used to determine whether to make corresponding classification recommendations for the user's conversation data through the classification agent. The matched conversation data is sent to each classification agent. After each classification agent receives the conversation data, it performs task detection of the classification recommendation task on the received conversation data to obtain its own detection results, and the detection results of each classification agent are input into the recommendation agent 103. The recommendation agent 103 generates conversations based on the input detection results and recommends the generated conversations to the user, thereby realizing conversation recommendations to the user through the cooperation between agents in the multi-agent system.

[0011] 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 belong to the user's privacy to a certain extent, therefore, if you want to collect the user's conversation data, service access data, user data and other related data, you can obtain the user's authorization before collecting the data, so that the operation of collecting data complies with relevant data management regulations. For example, data authorization can be performed when the user accesses the resource service for the first time, or during any time the user accesses the resource service, or when the user triggers the service identification. The specific method of data authorization can be to send a user data authorization reminder to the user, and the user can obtain the user's data authorization after confirming the reminder through an instruction, or the method of data authorization can also be to obtain the user's data authorization by signing a data authorization agreement.

[0012] One or more embodiments of a recommended processing method provided in this specification are as follows: Reference Figure 2 The recommendation processing method provided in this embodiment can be applied to a multi-agent system, and the method specifically includes steps S202 to S208.

[0013] Step S202: Acquire the conversation data of the user who is interacting with the resource service.

[0014] The resource service described in this embodiment refers to the resource-related service provided within the application, such as the resource service related to the security project provided by the application. Users can consult, inquire, purchase and / or sell the purchased security project by accessing the resource service. They can also apply for security of the security project or other processing related to the security project through the resource service. The security project can be a resource management security project that guarantees the user's resource management in the application, such as a financial security project.

[0015] The dialogue data refers to the data generated by the user's dialogue interaction during the access process of the resource service. Specifically, the dialogue data can be the dialogue data generated by the user's dialogue in the interaction area after the service identifier of the resource service is triggered.

[0016] Optionally, the dialogue interaction is conducted by triggering the service identifier of the resource service. The service identifier refers to the identifier configured by the resource service for interacting with the user. After the service identifier is triggered, the interaction area will be displayed, and the user can make corresponding consultations, inquiries and / or purchases in the interaction area. In addition, the resource service can also make corresponding resource recommendations or dialogue recommendations to the user in the interaction area. Optionally, the service identifier includes the interaction identifier of the virtual person or digital person configured by the resource service.

[0017] For example, when a user accesses the project service of an insurance project, the conversation data generated by the user's conversation interaction with the digital human configured by the project service is as follows: “Digital Man: Have you ever purchased health insurance before? User: No User: I have health insurance programs from other institutions Digital Human: Do you have any claims record for health insurance programs recently? User: No Digital Human: How much coverage do you want to purchase for health insurance? Users: 300,000".

[0018] In specific implementation, when a user accesses a resource service and conducts a conversation interaction, the conversation data obtained by the user in the conversation interaction with the resource service is obtained, and the method of recommending a conversation to the user in the process of the user accessing the resource service is recommended based on the conversation data.

[0019] Step S204, classifying and recommending matching of the conversation data by scheduling agents, and sending the conversation data to each classified agent after matching.

[0020] The scheduling agent in this embodiment refers to the agent responsible for scheduling in the multi-agent system; the classification agent refers to the agent that performs recommendation detection processing for each classification in the multi-agent system; wherein, after the scheduling agent performs classification recommendation matching, the matching results obtained include a match pass and a match fail, and a match pass indicates that further classification recommendation processing is required for the user's conversation data, and for the conversation data that passes the match, the conversation data that passes the match is sent to each classification agent respectively, so that each classification agent performs recommendation detection processing for the corresponding classification; Failed matching indicates that no further classification and recommendation processing is required for the user's conversation data. For conversation data that failed matching, a conversation response can be performed on the conversation data or no processing can be performed. For example, a conversation response can be performed by inputting the conversation data into a conversation generation model for conversation generation and recommending the generated conversation to the user.

[0021] In the specific implementation, based on the user's conversation data, the user's conversation data is classified and recommended and matched by scheduling the intelligent agent to determine whether the current conversation data needs further classification and recommendation processing. If the match is successful, the conversation data is sent to each classification intelligent agent, that is, the matched conversation data is sent to each classification intelligent agent, so that each classification intelligent agent can further classify and recommend the matched conversation data.

[0022] Specifically, in an optional implementation manner provided by this embodiment, classification recommendation detection is performed on the user's conversation data, including: Performing intent recognition on the user's conversation data to obtain the conversation intent, or performing semantic recognition on the user's conversation data to obtain the conversation semantics; Detect the matching degree between the dialogue intention or dialogue semantics and the classification recommendation tasks of each classification agent. If any matching degree is greater than a preset threshold, it is determined that the match is successful. In addition, if all the matching degrees are less than or equal to the preset threshold, it indicates that there is no need to perform further classification and recommendation processing on the user's conversation data. In this case, a conversation response to the conversation data can be performed or no processing can be performed. For example, a conversation response to the conversation data can be performed by inputting the conversation data into a conversation generation model for conversation generation and recommending the generated conversation to the user.

[0023] It should be noted that the process of classifying, recommending and matching the user's conversation data can be achieved through a large language model. For example, the user's conversation data can be classified, recommended and matched through a large language model to obtain a matching result, and the matching result obtained includes a passed match or a failed match. Specifically, a matching text containing conversation data and a classification recommendation matching field can be created, and the matching text containing conversation data and the classification recommendation matching field can be input into the large language model for classification recommendation matching to output a matching result. In addition, the large language model can be fine-tuned or trained in advance so that the large language model has the ability to perform classification recommendation matching and output matching results after fine-tuning or training, so that the conversation data can be input into the large language model for classification recommendation matching to output a matching result.

[0024] Similarly, the above-mentioned intention recognition of the user's conversation data to obtain the conversation intention, and the detection of the matching degree of the conversation intention with the classification recommendation task of each classification agent can also be implemented by using a large language model, such as creating a text containing conversation data and a processing field for conversation intention recognition and matching degree calculation, inputting the text containing the conversation data and the processing field into the large language model to perform intention recognition and match the conversation intention with the classification recommendation task of each classification agent, and outputting the matching degree; Alternatively, the above-mentioned semantic recognition of the user's conversation data to obtain the conversation semantics and detect the matching degree of the conversation semantics with the classification recommendation tasks of each classification agent can also be implemented using a large language model. For example, a text containing the conversation data and processing fields for performing conversation semantic recognition and matching degree calculation is created, and semantic recognition is performed by inputting the text containing the conversation data and the processing fields into the large language model, and the matching degree of the conversation semantics with the classification recommendation tasks of each classification agent is detected, and the matching degree is output.

[0025] Step S206, the classification agents perform task detection of the classification recommendation tasks on the dialogue data to obtain detection results.

[0026] In this embodiment, multiple classification agents are set up in a multi-agent system, and each classification agent is used to perform recommendation detection processing on the user's conversation data in the corresponding classification. Specifically, the recommendation processing performed by each classification agent includes task detection of the respective classification recommendation tasks on the user's conversation data, that is, each classification agent realizes recommendation detection processing on the user's conversation data by performing task detection of the respective classification recommendation tasks on the user's conversation data.

[0027] Optionally, the classification agents correspond one-to-one to the classification recommendation tasks, and each classification agent performs task execution detection on its corresponding classification recommendation task.

[0028] Among them, the classification recommendation task refers to a dialogue recommendation task in which a classification agent performs dialogue recommendation on the user's dialogue data under a single classification (dimension). The task detection of the classification recommendation task on the user's dialogue data refers to detecting whether the dialogue recommendation of the classification recommendation task is performed 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 types of task detection results for the classification recommendation task, one is to make a recommendation, and the other is not to make a recommendation.

[0029] Optionally, the classified recommendation tasks include: a payment change recommendation task for recommending to the user a payment method for increasing the payment period of the insurance item, and / or an increase recommendation task for recommending to the user an increase in the insurance amount of the insurance item.

[0030] For example, when a user accesses a guarantee project in a resource service, the user clicks on the service logo of the resource service to interact with the digital human of the resource service. Based on the conversation data generated by the conversation between the user and the digital human, the digital human will determine whether it is necessary to make a conversation recommendation to the user. Here, the classification recommendation task corresponding to the classification agent A in the multi-agent system is: recommending to the user to increase the payment cycle of the guarantee project, and the classification recommendation task is specifically 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. The classification recommendation task corresponding to the classification agent B in the multi-agent system is: to recommend to the user to increase the insurance amount of the insurance item. The classification recommendation task is to recommend to the user to increase the insurance amount of the insurance item currently selected by the user (for example, 100,000) to the preset insurance amount (for example, 300,000).

[0031] In the specific implementation, each classification agent performs task detection of each classification recommendation task on the user's conversation data. Specifically, each classification agent performs task detection of the classification recommendation task corresponding to each classification agent on the user's conversation data to obtain the detection results of each classification agent.

[0032] Specifically, in the process of performing task detection of each category recommendation task on the user's conversation data, the task detection of each category recommendation task may also be performed through a large language model. In an optional implementation manner provided in this embodiment, performing task detection of each category recommendation task on the user's conversation data to obtain a detection result includes: Generate task detection texts for each task based on the dialogue data and the task detection templates for each classification recommendation task; The task detection texts are respectively input into the large language model to perform task execution detection of the classification recommendation task recorded in the task detection text to obtain the detection results.

[0033] Among them, the task detection template of the classification recommendation task refers to the description text used to instruct the large language model to perform task detection for the classification recommendation task; for example, the description sentence in the task detection template of the classification recommendation task corresponding to the classification agent A in the multi-agent system is: "You are an AI assistant familiar with insurance sales. Please judge whether it is the right time to recommend to the user that the quarterly payment method or the annual payment method of the insurance project has an advantage over the monthly payment method after the user's last conversation?"; For another example, the description statement in the task detection template of the classification recommendation task corresponding to the classification agent B in the multi-agent system is: "You are an AI assistant familiar with insurance sales. Please determine whether it is time to recommend that the user increase the insurance amount of the currently selected insurance item to the preset insurance amount after the user's last sentence of dialogue?"

[0034] In actual applications, in order to improve the effectiveness and accuracy of conversation recommendations to users based on their conversation data, and also to improve the conversion rate of conversation recommendations to users, it is also possible to set execution conditional statements for each category recommendation task in the task detection template of each category recommendation task. That is, by setting execution conditional 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 the execution conditional statements, thereby making the timing of conversation recommendations to users more adapted to user needs, which helps to improve the conversion rate of conversation recommendations to users.

[0035] Optionally, at least one execution condition statement of each classification recommendation task is recorded in the task detection template of each classification recommendation task, that is, at least one execution condition statement of each classification recommendation task is recorded in the task detection template of the classification recommendation task.

[0036] For example, the execution conditional statements recorded in the task detection template of the classification recommendation task corresponding to the classification agent A in the multi-agent system include: “1. When the user’s last sentence is asking about the payment method, such as annual payment or monthly payment, this opportunity can be triggered; 2. This opportunity can be triggered when the user's last sentence indicates the intention to pay monthly; 3. In the previous conversation with the user, the user has been recommended to pay annually, so this opportunity cannot be triggered; 4. The user has already asked about the payment method during the conversation, so this opportunity cannot be triggered; 5. It can be clearly determined in the dialogue that the current insurance product is not a designated health insurance product and cannot trigger this opportunity. ".

[0037] Specifically, based on the fact that at least one execution conditional statement of each classification recommendation task is recorded in the task detection template of the classification recommendation task, during the process of task execution detection of the classification recommendation task recorded in the task detection text by the large language model, a binary classification detection can be performed on each execution conditional statement, and finally the detection of the classification recommendation task is determined according to the detection result of each binary classification detection.

[0038] In an optional implementation manner provided in this embodiment, performing task execution detection of a classification recommendation task of a task detection text record includes: Reading a description statement of a classification recommendation task and at least one execution conditional statement recorded in a classification detection text; A binary classification test is performed on whether the dialogue data triggers the description statement and each execution condition statement, and the test result is determined according to the binary classification test result.

[0039] In the present embodiment, in the implementation method of performing task detection for each classification recommendation task through a large language model, before the large language model performs task detection for each classification recommendation task, the large language model capable of performing task detection for each classification recommendation task can also be obtained by fine-tuning the pre-trained model. Specifically, in an optional implementation method provided by the present embodiment, the large language model is obtained by training in the following manner: creating a training text containing task records of classification recommendation tasks for each classification agent, and inputting the training text into the pre-trained model, and the pre-trained model performs model fine-tuning according to the training text to obtain the large language model.

[0040] A pre-trained model refers to a natural language model obtained through pre-training. The pre-trained model can adopt a neural network architecture containing a large number of parameters. The pre-trained model can adopt a pre-trained large language model (LLM), such as chatGPT (chat Generative Pre-trained Transformer) and other open source large language models.

[0041] It should be noted that the classification recommendation task can be a binary classification recommendation task, that is, the task detection result of each classification recommendation task is one of the two detection results of recommendation or non-recommendation. Multiple binary classification recommendation tasks can be obtained by splitting the multi-label classification recommendation task, or multiple binary classification recommendation tasks can be obtained by splitting the multi-classification recommendation task. Here, the multi-label classification recommendation task refers to a dialogue recommendation task that classifies the user's conversation data under multiple categories (dimensions) and finally determines one or more classification labels for dialogue recommendation; the multi-classification recommendation task refers to a dialogue recommendation task that classifies the user's conversation data under multiple categories and finally determines one category for dialogue recommendation; based on this, the user will be given a recommendation. After the multi-label classification recommendation task for user dialogue recommendation or the multi-classification recommendation task is split into multiple binary classification recommendation tasks, since the probability of the triggering timing of each binary classification recommendation task is low, that is, the probability of each binary classification recommendation task being triggered is relatively low, therefore, if the pre-trained model is fine-tuned to obtain a large language model by adopting the multi-label classification recommendation task or the multi-classification recommendation task, the number of training samples required for targeted training of each classification is large, while the number of training samples required for fine-tuning the pre-trained model to obtain a large language model by adopting the binary classification recommendation task is greatly reduced, and only corresponding training samples need to be constructed for each binary classification recommendation task.

[0042] Step S208: input each detection result obtained into the recommendation agent to generate a dialogue and make a dialogue recommendation to the user.

[0043] In the above, each classification agent performs task detection of each classification recommendation task on the user's dialogue data to obtain the detection result of each classification recommendation task, that is, the detection result of each classification agent is obtained. Based on this, the detection result of each classification recommendation task is input into the recommendation agent, and the dialogue is generated by the recommendation agent and the dialogue recommendation is made to the user.

[0044] As described above, the detection results of each classification agent include either making a recommendation or not making a recommendation. Among the multiple detection results of multiple classification agents, there may be a case where the detection results of one or more classification agents are for recommendation, or there may be a case where the detection results of all classification agents are for not recommending. In this case, a dialogue response to the dialogue data may be performed or no processing may be performed. For example, a dialogue response to the dialogue data may be performed by inputting the dialogue data into a dialogue generation model for dialogue generation and recommending the generated dialogue to the user.

[0045] Specifically, in the process of the recommendation agent generating a dialogue and recommending a dialogue to a user, for a situation where the detection result of only one classification agent among multiple detection results of multiple classification agents is for recommendation, in an optional implementation manner provided by this embodiment, generating a dialogue and recommending a dialogue to a user includes: If each detection result includes only one detection result for recommendation, a recommendation dialog for the classification recommendation task corresponding to the detection result for recommendation is generated, and a recommendation node of the recommendation dialog is determined, so as to send a recommendation dialog to the user according to the recommendation node; or, If each detection result includes only one detection result for recommendation, a recommendation dialogue for the classification recommendation task corresponding to the detection result for recommendation is generated, and the combined recommendation dialogue is displayed in the dialogue area after the service identifier of the resource service is triggered.

[0046] In addition, for the case where at least two of the detection results of the multiple classification agents are for recommendation, in an optional implementation manner provided by this embodiment, a dialogue is generated and a dialogue recommendation is made to the user, including: If each detection result includes at least two detection results for recommendation, a merged recommendation dialogue is generated according to the dialogue data and at least two classification recommendation tasks corresponding to the at least two detection results for recommendation, and the merged recommendation dialogue is displayed in the dialogue area after the service identifier of the resource service is triggered; or, If each detection result includes at least two detection results for recommendation, a merged recommendation dialogue is generated according to the dialogue data and at least two classification recommendation tasks corresponding to the at least two detection results for recommendation, and a recommendation node of the merged recommendation dialogue is determined to send the merged recommendation dialogue to the user according to the recommendation node.

[0047] Optionally, the combined recommendation dialogue or the recommendation dialogue generated by the recommendation agent is displayed in the interactive area displayed after the service identifier is triggered.

[0048] During the specific execution process, the above-mentioned recommendation dialogue for generating the classification recommendation task corresponding to the detection result for recommendation can be implemented 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 and the dialogue generation sentence corresponding to the detection result can be input into the dialogue generation model for dialogue generation to obtain the recommendation dialogue; similarly, the above-mentioned generation of the merged recommendation dialogue based on the dialogue data and at least two classification recommendation tasks corresponding to at least two detection results for recommendation can be implemented through a dialogue generation model. Specifically, the dialogue generation text containing at least two classification recommendation tasks and the merged dialogue generation sentence can be input into the dialogue generation model for dialogue generation to obtain the merged recommendation dialogue.

[0049] It should be pointed out that the above implementation method provides multiple machine learning models, specifically including: a large language model that inputs task detection text to perform task execution detection, a dialogue generation model that inputs dialogue data to generate dialogues, and a dialogue generation model that inputs dialogue generation text to generate dialogues, wherein the dialogue generation model that inputs dialogue data to generate dialogues and the dialogue generation model that inputs dialogue generation text to generate dialogues can be the same dialogue generation model, and the dialogue generation model can be a large language model. In this case, the large language model that performs dialogue generation and the large language model that performs task execution detection can be the same large language model or different large language models.

[0050] In one or more embodiments, the display of the service identification may be determined based on the user's service access data and / or user data, and may also be determined based on other data, such as service data of the resource service and / or resource data of the resource, or, whether to display the service identification may also be determined based on at least one of the service access data and the user data and at least one of the service data and the resource data of the resource; Specifically, the display of the service identifier may correspond to the resource service or the service page of the resource service. For example, the service identifier may be displayed on a specific service page of the resource service, and the service identifier may not be displayed on other service pages other than the specific service page of the resource service. Specifically, the specific service page for displaying the service identifier may be determined by inputting service access data, user data, service data of the resource service and / or resource data of the resource into a pre-trained identifier decision model or a large language model; In addition, the display of the service identifier may also correspond to the resources in the resource service. For example, the service identifier may be displayed on the resource page of a specific resource in the resource service, and the service identifier may not be displayed on the resource page of other resources other than the specific resource. Specifically, the specific resource displaying the service identifier may be determined by inputting the user's service access data, user data, service data of the resource service and / or resource data of the resource into a pre-trained identification decision model or a large language model, or the resource page of the specific resource displaying the service identifier may be determined by inputting the service access data, user data, service data of the resource service and / or resource data of the resource into a pre-trained identification decision model or a large language model; or The display of the logo can also correspond to the user, such as displaying the service logo during the access process of a specific user. Specifically, it can be determined whether to display the service logo for the user by inputting the user's service access data, user data, service data of the resource service and / or resource data of the resource into a pre-trained logo decision model or a large language model; alternatively, the display of the service logo can also correspond to at least two of the resource service, the service page of the resource service, the resource and the user. Similarly, it can be determined whether to display the result for the service logo by inputting the user's service access data, user data, service data of the resource service and / or resource data of the resource into a pre-trained logo decision model or a large language model.

[0051] In summary, the recommendation processing method provided in this embodiment obtains the conversation data obtained by the user in the conversation interaction with the resource service on the basis of the user accessing the resource service and conducting the conversation interaction, and classifies and recommends the user's conversation data through the scheduling agent to determine whether the current conversation data needs to be further classified and recommended. If the match is successful, the conversation data is sent to each classification agent. After each classification agent receives the sent conversation data, in the process of performing task detection of each classification recommendation task on the user's conversation data, specifically, each classification agent performs task detection of the classification recommendation task corresponding to each classification agent on the user's conversation data to obtain the detection result of each classification agent. Further, the detection result of each classification recommendation task is input into the recommendation agent, and the recommendation agent generates a conversation and makes a conversation recommendation to the user. In this way, the conversation data of the user is made to cooperate with each other in the multi-agent system, thereby making a conversation recommendation to the user based on the user's conversation data, thereby improving the effectiveness and accuracy of the conversation recommendation to the user, enabling the timing of the conversation recommendation to the user to be more adapted to the user's needs, and helping to improve the recommendation conversion rate of the conversation recommendation to the user.

[0052] The following takes the application of a recommendation processing method provided in this embodiment in a resource service scenario as an example, combined with Figure 3 , for further explanation of the recommended processing method provided in this embodiment, see Figure 3, the recommended processing method applied to the resource service scenario specifically includes the following steps.

[0053] Step S302: Acquire the conversation data of the user who is interacting with the resource service.

[0054] Step S304, classify and recommend matches for the dialogue data through the scheduling agent in the multi-agent system to obtain matching results.

[0055] Step S306: If the matching result is a successful match, the dialogue data is sent to the first classification agent and the second classification agent in the multi-agent system.

[0056] Step S308: The first classification agent obtains a first detection result by performing task detection of the first classification recommendation task on the dialogue data.

[0057] Step S310, obtaining a second detection result by the second classification agent performing task detection of the second classification recommendation task on the dialogue data.

[0058] Step S312, input the first detection result and the second detection result into the recommended agent in the multi-agent system.

[0059] Step S314: If the first detection result or the second detection result is for recommendation, a recommendation dialogue of the corresponding first category recommendation task or second category recommendation task is generated by the recommendation agent and sent to the user.

[0060] Step S316: If both the first detection result and the second detection result are for recommendation, a combined recommendation dialogue of the first category recommendation task and the second category recommendation task is generated by the recommendation agent and sent to the user.

[0061] It should be noted that any one of steps S302 to S316 or any combination of multiple steps can be combined with any one of steps S202 to S208 to form a new implementation method according to the needs of implementation deployment; in addition, according to the needs of actual deployment, any one or multiple technical features can be selected from steps S302 to S316 and combined with any one or multiple technical features provided by steps S202 to S208 to form a new implementation method; or, any one or multiple technical features in steps S302 to S316 can be replaced with any one or multiple technical features provided by steps S202 to S208 to form a new implementation method according to the needs of actual deployment, which will not be repeated here.

[0062] An example of a recommended processing device provided in this specification is as follows: In the above-mentioned embodiment, a recommendation processing method is provided, and correspondingly, a recommendation processing device is also provided, which will be described below with reference to the accompanying drawings.

[0063] Reference Figure 4 , which shows a schematic diagram of an embodiment of a recommendation processing device provided in this embodiment.

[0064] Since the device embodiment corresponds to the method embodiment, the description is relatively simple, and the relevant parts can refer to the corresponding description of the method embodiment provided above. The device embodiment described below is only illustrative.

[0065] This embodiment provides a recommendation processing device, the device comprising: The conversation acquisition module 402 is configured to acquire conversation data of users who are interacting with each other in the resource service; The recommendation matching module 404 is configured to classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the match is passed; A classification detection module 406 is configured to perform task detection of each classification recommendation task on the dialogue data through each classification agent to obtain a detection result; The dialogue generation module 408 is configured to input the obtained detection results into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0066] An example of a recommended processing device provided in this specification is as follows: Corresponding to the recommendation processing method described above, based on the same technical concept, one or more embodiments of this specification further provide a recommendation processing device, which is used to execute the recommendation processing method provided above. Figure 5 A schematic diagram of the structure of a recommended processing device provided for one or more embodiments of this specification.

[0067] This embodiment provides a recommendation processing device, including: like Figure 5As shown, the recommendation processing device may have relatively large differences due to different configurations or performances, and may include one or more processors 501 and memory 502, and the memory 502 may store one or more storage applications or data. Among them, the memory 502 can be a short-term storage or a persistent storage. The application stored in the memory 502 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the recommendation processing device. Furthermore, the processor 501 can be configured to communicate with the memory 502 and execute a series of computer executable instructions in the memory 502 on the recommendation processing device. The recommendation 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.

[0068] In a specific embodiment, the recommendation 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 recommendation processing device, and the one or more programs are configured to be executed by one or more processors, including computer executable instructions for performing the following: Obtaining conversation data of users who are interacting with the resource service; Classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the matching is successful; The classification agents perform task detection of each classification recommendation task on the dialogue data to obtain a detection result; The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0069] An embodiment of a computer-readable storage medium provided in this specification is as follows: Corresponding to the recommended processing method described above, based on the same technical concept, one or more embodiments of this specification also provide a computer-readable storage medium.

[0070] The computer-readable storage medium provided in this embodiment is used to store computer-executable instructions, and the computer-executable instructions implement the following process when executed: Obtaining conversation data of users who are interacting with the resource service; Classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the matching is successful; The classification agents perform task detection of each classification recommendation task on the dialogue data to obtain a detection result; The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0071] It should be noted that an embodiment of a computer-readable storage medium in this specification and an embodiment of a recommended processing method in this specification are based on the same inventive concept, so the specific implementation of this embodiment can refer to the implementation of the aforementioned corresponding method, and the repeated parts will not be repeated.

[0072] An embodiment of a computer program product provided in this specification is as follows: Corresponding to the recommendation processing method described above, based on the same technical concept, one or more embodiments of this specification also provide a computer program product.

[0073] A computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the following steps: Obtaining conversation data of users who are interacting with the resource service; Classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the matching is successful; The classification agents perform task detection of each classification recommendation task on the dialogue data to obtain a detection result; The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

[0074] It should be noted that an embodiment of a computer program product in this specification and an embodiment of a recommended processing method in this specification are based on the same inventive concept, so the specific implementation of this embodiment can refer to the implementation of the aforementioned corresponding method, and the repeated parts will not be repeated.

[0075] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. For example, the device embodiment, equipment embodiment, computer-readable storage medium embodiment, and computer program product embodiment are similar to the method embodiment, so the description is relatively simple. To read the relevant contents of the device embodiment, equipment embodiment, computer-readable storage medium embodiment, and computer program product embodiment, please refer to the partial description of the method embodiment.

[0076] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] In the 1930s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to circuit structures such as diodes, transistors, switches, etc.) or software improvements (improvements to method flows). However, with the development of technology, many of today's method flow improvements can be regarded as direct improvements to hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using hardware entity modules. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask chip manufacturers to design and produce dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, 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, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0078] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0079] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0080] For the convenience of description, the above devices are described in terms of functions and are divided into various units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0081] It should be understood by those skilled in the art that one or more embodiments of the present specification may be provided as a method, system or computer program product. Therefore, one or more embodiments of the present specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present specification may take the form of a computer program product implemented 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.

[0082] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0084] These computer program instructions may also be loaded onto a computer or other programmable processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0085] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0086] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0087] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. 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 technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0088] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, the elements defined by the sentence "includes at least one ..." do not exclude the presence of other identical elements in the process, method, commodity or device including the elements.

[0089] One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0090] The above description is only an embodiment of this document and is not intended to limit this document. For those skilled in the art, this document may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this document should be included in the scope of the claims of this document.

Claims

1. A recommendation processing method, applied to a multi-agent system, comprising: Obtaining conversation data of users who are interacting with the resource service; Classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the matching is successful; The classification agents perform task detection of each classification recommendation task on the dialogue data to obtain a detection result; The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

2. According to the recommendation processing method of claim 1, the classification agent corresponds to the classification recommendation task one by one, and each classification agent performs task execution detection for the classification recommendation task corresponding to the classification agent; The task detection includes a binary classification detection of whether the dialogue data triggers a task execution condition of the classification recommendation task.

3. The recommendation processing method according to claim 2, wherein the classification recommendation task comprises: A payment change recommendation task is recommended to increase the payment period of the insurance item and / or an increase recommendation task is recommended to increase the insurance amount of the insurance item.

4. The recommendation processing method according to claim 1, wherein the step of performing task detection on the conversation data for each classification recommendation task to obtain a detection result comprises: Generate each task detection text according to the dialogue data and the task detection template of each classified recommended task; The task detection texts are respectively input into the large language model to perform task execution detection of the classification recommendation task recorded in the task detection text to obtain the detection result.

5. The recommendation processing method according to claim 4, wherein the task execution detection of the classified recommendation task of the task detection text record comprises: Reading a description statement of a classification recommendation task and at least one execution conditional statement recorded in a classification detection text; A binary classification detection is performed on whether the dialog data triggers the description statement and each execution condition statement, and the detection result is determined according to the binary classification detection result.

6. According to the recommendation processing method of claim 4, the large language model is trained in the following manner: A training text containing task records of the classification recommendation tasks of the classification agents is created, and the training text is input into a pre-training model, and the pre-training model is fine-tuned according to the training text to obtain the large language model.

7. The recommendation processing method according to claim 1, wherein the step of generating a dialog and making a dialog recommendation to the user comprises: If the detection results include only one detection result for recommendation, a recommendation dialogue for the classification recommendation task corresponding to the detection result for recommendation is generated, and a recommendation node of the recommendation dialogue is determined to send the recommendation dialogue to the user according to the recommendation node.

8. The recommendation processing method according to claim 1, wherein generating a dialogue and making a dialogue recommendation to the user comprises: If the detection results include at least two detection results for recommendation, a merged recommendation dialogue is generated based on the dialogue data and at least two classification recommendation tasks corresponding to the at least two detection results for recommendation, and the merged recommendation dialogue is displayed in the dialogue area after the service identifier of the resource service is triggered.

9. The recommendation processing method according to claim 1, wherein the classifying and recommending detection of the conversation data comprises: Performing intent recognition on the conversation data to obtain conversation intent, or performing semantic recognition on the conversation data to obtain conversation semantics; The matching degree between the dialogue intention or the dialogue semantics and the classification recommendation tasks of each classification agent is detected. If any matching degree is greater than a preset threshold, it is determined that the matching is successful.

10. The recommendation processing method according to claim 9, after the classifying and recommending matching operation of the dialogue data by scheduling an agent, further comprising: If the matching degree is less than or equal to the preset threshold, the dialogue data is input into a dialogue generation model to generate a dialogue, and the generated dialogue is recommended to the user.

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

12. A recommendation processing device, running on a multi-agent system, comprising: A conversation acquisition module, configured to acquire conversation data of users who are interacting with each other in a conversation in a resource service; A recommendation matching module is configured to classify and recommend matching of the conversation data by scheduling agents, and send the conversation data to each classification agent after matching; A classification detection module is configured to obtain a detection result by performing task detection of each classification recommendation task on the dialogue data through each classification agent; The dialogue generation module is configured to input the obtained detection results into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

13. A recommendation processing device, comprising: processor; and a memory configured to store computer executable instructions that, when executed, cause the processor to: Obtaining conversation data of users who are interacting with the resource service; Classify and recommend the conversation data by scheduling the intelligent agent, and send the conversation data to each classified intelligent agent after the matching is successful; The classification agents perform task detection of each classification recommendation task on the dialogue data to obtain a detection result; The obtained detection results are input into the recommendation agent to generate dialogue and make dialogue recommendations to the user.

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

Images