Credit card recommendation method and device, equipment, medium and program product

Abstract

The invention provides a credit card recommendation method which can be applied to the technical field of artificial intelligence, and particularly relates to application of a large language model in the field of financial science and technology. The credit card recommendation method comprises the following steps: in response to receiving consultation information input by a user, analyzing the consultation information by using a large language model to obtain dominant information and invisible requirements in the consultation information; extracting attribute information of the user according to the dominant information and the invisible demand, and updating a user portrait state according to the attribute information; in response to the condition that the user portrait state meets the recommendation condition, searching at least one target credit card matched with a target portrait from a pre-constructed multi-source knowledge base according to the target portrait corresponding to the current user portrait state; and according to the at least one target credit card and the target portrait, utilizing a large language model to generate a recommendation reason, and recommending the at least one target credit card and the recommendation reason to the user. The invention further provides a credit card recommendation device and equipment, a medium and a program product.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to application of large language models in the field of financial technology, and more particularly to a credit card recommendation method, device, equipment, medium and program product. BACKGROUND

[0002] The existing credit card recommendation method relies on single rules and fixed recommendation logic, which leads to the need for high-frequency reconstruction or deployment when the system adapts to changes in business rules and data sources, which seriously affects the resource utilization of the system. In addition, the existing credit card recommendation method is limited in user data processing and analysis capabilities, making it difficult to deeply mine complex user portrait features and implicit needs from multi-source and dynamic user interaction data, which leads to low computational efficiency of the recommendation algorithm and makes it difficult to generate high-precision personalized recommendation results, affecting the reliability of the system in processing complex information. SUMMARY

[0003] In view of the above problems, the present application provides a credit card recommendation method, device, equipment, medium and program product for improving system operation efficiency and response capability.

[0004] According to a first aspect of the present application, a credit card recommendation method is provided, comprising: in response to receiving user input consultation information, analyzing the consultation information using a large language model to obtain explicit information and implicit needs in the consultation information; extracting attribute information of the user according to the explicit information and the implicit needs, and updating the user portrait state according to the attribute information; in response to the user portrait state satisfying a recommendation condition, retrieving at least one target credit card matching a target portrait corresponding to the current user portrait state from a pre-constructed multi-source knowledge base according to the target portrait; generating a recommendation reason using the large language model according to the at least one target credit card and the target portrait, and recommending the at least one target credit card and the recommendation reason to the user.

[0005] According to an embodiment of the present application, before responding to receiving user input consultation information, analyzing the consultation information using a large language model to obtain explicit information and implicit needs in the consultation information, it includes: in response to the user initiating a consultation request, receiving the consultation request using a pre-configured multi-modal input interface; wherein the multi-modal input interface is configured to receive multiple modal data in the form of text, voice, image, and document.

[0006] According to an embodiment of the present application, in response to receiving the consultation information input by the user, the consultation information is parsed using a large language model to obtain explicit information and implicit needs in the consultation information, including: in response to receiving the consultation information input by the user, the named entities in the consultation information are identified using the large language model, and the named entities are subjected to intent recognition to obtain the explicit information; according to the named entities, the potential needs of the user are predicted using prompt words to obtain the implicit needs, wherein the prompt words are used to guide the large language model to perform a reasoning task according to a specified path.

[0007] According to an embodiment of the present application, according to the explicit information and the implicit needs, attribute information of the user is extracted, including: according to the explicit information and the implicit needs, authorization of the user to extract income information is obtained, and after obtaining the authorization of the user to extract the income information, income information of the user is extracted; in response to successful extraction of the income information, authorization of the user to extract occupation information is obtained, and after obtaining the authorization of the user to extract the occupation information, occupation information of the user is extracted; in response to successful extraction of the occupation information, authorization of the user to extract consumption habits is obtained, and after obtaining the authorization of the user to extract the consumption habits, consumption habits of the user are extracted.

[0008] According to an embodiment of the present application, according to the attribute information, a user portrait state is updated, including: obtaining authorization of the user to extract identity information, and after obtaining the authorization of the user to extract the identity information, identity information of the user is extracted; according to the identity information, an initial portrait of the user is constructed; based on the initial portrait, the income information, the occupation information and the consumption habits are used to update a state of the initial portrait to obtain a target portrait.

[0009] According to an embodiment of the present application, the user portrait state is updated according to the attribute information, and further includes: in response to the target portrait not satisfying a recommendation condition, authorization of the user to extract historical behavior data is obtained, and after obtaining the authorization of the user to extract the historical behavior data, historical behavior data of the user is extracted; according to the historical behavior data, the target portrait is updated.

[0010] According to an embodiment of the present application, in response to the user portrait state satisfying a recommendation condition, at least one target credit card matched with a target portrait corresponding to a current user portrait state is retrieved from a pre-constructed multi-source knowledge base, including: a matching degree between the target portrait and all credit cards that can be retrieved in the multi-source knowledge base is calculated; according to the matching degree, the retrieved credit cards are subjected to sorting processing; based on the credit cards subjected to the sorting processing, the top k credit cards with the highest matching degrees are selected as the target credit cards.

[0011] According to an embodiment of the present application, the recommendation reason is generated by using the large language model according to the at least one target credit card and the target portrait, including: filling the information of the at least one target credit card and the information of the target portrait into a preset template, wherein the preset template matches the target portrait; and guiding the large language model to generate the recommendation reason corresponding to the preset template according to the preset template.

[0012] The second aspect of the present application provides a credit card recommendation device, including: an analysis module configured to analyze the consultation information input by a user by using a large language model to obtain explicit information and implicit requirements in the consultation information; an extraction module configured to extract attribute information of the user according to the explicit information and the implicit requirements, and update a user portrait state according to the attribute information; a retrieval module configured to retrieve at least one target credit card matching a target portrait from a pre-constructed multi-source knowledge base according to the target portrait corresponding to the current user portrait state in response to the user portrait state satisfying a recommendation condition; and a recommendation module configured to generate a recommendation reason by using a large language model according to the at least one target credit card and the target portrait, and recommend the at least one target credit card and the recommendation reason to the user.

[0013] The third aspect of the present application provides an electronic device, including: one or more processors; and a memory configured to store one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method.

[0014] The fourth aspect of the present application further provides a computer-readable storage medium having a computer program or instructions stored thereon, wherein the computer program or instructions are executed by a processor to implement the steps of the method.

[0015] The fifth aspect of the present application further provides a computer program product including a computer program or instructions, wherein the computer program or instructions are executed by a processor to implement the steps of the method. BRIEF DESCRIPTION OF DRAWINGS

[0016] The above content and other purposes, features and advantages of the present application will be more apparent from the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0017] Figure 1 An application scenario diagram of a credit card recommendation method according to an embodiment of the present application is schematically shown;

[0018] Figure 2 A flowchart of a credit card recommendation method according to an embodiment of the present application is schematically shown;

[0019] Figure 3 A principle diagram of input data received by a multi-modal input interface according to an embodiment of the present application is schematically shown.

[0020] Figure 4 a schematic diagram illustrating a principle of a credit card recommendation method according to an embodiment of the present application is shown;

[0021] Figure 5 a structural block diagram of a credit card recommendation apparatus according to an embodiment of the present application is shown; and

[0022] Figure 6 a block diagram of an electronic device suitable for implementing a credit card recommendation method according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0023] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. It is to be understood, however, that the description is merely exemplary of the present application, and is not intended to limit the scope of the present application. In the following detailed description of the embodiments of the present application, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without these specific details. In other instances, well-known structures and functions have not been described in detail in order to avoid obscuring aspects of the present application.

[0024] The terms used herein are merely used to describe specific embodiments, and are not intended to limit the present application. The terms "include" and "have" and the like used herein indicate the presence of the described features, steps, operations, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

[0025] All terms used herein, including technical and scientific terms, have the same meanings as those generally understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having meanings consistent with the context of the present description, and should not be interpreted in an idealized or overly formal manner.

[0026] In the case of using expressions similar to "at least one of A, B, and C, etc.", it is generally to be interpreted as including one or more of the same. For example, "a system having at least one of A, B, and C" should be interpreted as including a system having A alone, a system having B alone, a system having C alone, a system having A and B together, a system having A and C together, a system having B and C together, and / or a system having A, B, and C together, etc.

[0027] It should be noted that the credit card recommendation method and apparatus of the present application can be used for applications in the field of financial technology in the credit card recommendation, and can also be used for applications in any field other than the field of financial technology in the credit card recommendation. The application field of the credit card recommendation method and apparatus of the present application is not limited.

[0028] In the technical solutions of the present application, the user information (including but not limited to user personal information, user image information, user device information such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards, necessary security measures are taken, do not violate public order and good customs, and appropriate operation portals are provided for users to choose authorization or refusal.

[0029] In the scenario of making automated decisions using personal information, the method, device and system provided by the embodiments of the present application all provide corresponding operation portals for users to choose to agree or refuse the automated decision result; if the user chooses to refuse, the expert decision process is entered. The expression "automated decision" here refers to the activity of automatically analyzing, evaluating the behavior habits, interests and hobbies or economic, health, credit status of individuals, etc. by computer programs and making decisions. The expression "expert decision" here refers to the activity of making decisions by personnel who are engaged in a certain field of work, have special experience, knowledge and skills and reach a certain professional level.

[0030] The embodiment of the present application provides a credit card recommendation method, comprising: in response to receiving user input consultation information, analyzing the consultation information by using a large language model to obtain explicit information and hidden demand in the consultation information; according to the explicit information and the hidden demand, extracting attribute information of the user, and updating a user portrait state according to the attribute information; in response to the user portrait state satisfying a recommendation condition, retrieving at least one target credit card matched with a target portrait corresponding to the current user portrait state from a pre-constructed multi-source knowledge base according to the target portrait; generating a recommendation reason by using the large language model according to the at least one target credit card and the target portrait, and recommending the at least one target credit card and the recommendation reason to the user.

[0031] Figure 1 The application scenario diagram of the credit card recommendation method according to the embodiment of the present application is schematically shown.

[0032] As Figure 1 shown, the application scenario 100 according to the embodiment can include a first terminal device 101, a second terminal device 102, and a third terminal device 103. The network 104 is a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0033] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0034] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smartphones, tablet computers, laptop computers, desktop computers, etc.

[0035] The server 105 can be a server providing various services, such as a background management server providing support for websites browsed by the user using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (only as examples). The background management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data, etc. obtained or generated according to user requests) to the terminal device.

[0036] It should be noted that the credit card recommendation method provided by the embodiments of the present application can generally be executed by the server 105. Correspondingly, the credit card recommendation device provided by the embodiments of the present application can generally be arranged in the server 105. The credit card recommendation method provided by the embodiments of the present application can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Correspondingly, the credit card recommendation device provided by the embodiments of the present application can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0037] It should be understood that Figure 1 The number of terminal devices, networks, and servers in

[0038] The credit card recommendation method according to the embodiments of the present application will be described in detail below based on the scenarios described in Figure 1 Figures 2-4

[0039] Figure 2 The flowchart of the credit card recommendation method according to the embodiments of the present application is schematically shown.

[0040] As​​Figure 2 As shown, the credit card recommendation method of this embodiment includes operations S210-S240, which can be performed by an intelligent agent.

[0041] In operation S210, in response to receiving user input consultation information, the consultation information is parsed using a large language model to obtain explicit information and implicit requirements in the consultation information.

[0042] In operation S220, attribute information of the user is extracted according to the explicit information and the implicit requirements, and the user portrait state is updated according to the attribute information.

[0043] In operation S230, in response to the user portrait state satisfying a recommendation condition, at least one target credit card matching a target portrait corresponding to the current user portrait state is retrieved from a pre-constructed multi-source knowledge base.

[0044] In operation S240, a recommendation reason is generated using a large language model according to the at least one target credit card and the target portrait, and the at least one target credit card and the recommendation reason are recommended to the user.

[0045] In the embodiments of the present application, a large language model (LLM) is introduced to deeply mine the user's consultation information and fully extract the explicit information and implicit requirements therein. On this basis, a user portrait dynamic updating mechanism + multi-source database retrieval + LLM recommendation reason generation workflow engine is established to realize multi-dimensional information adjustment and personalized recommendation.

[0046] For example, the user says through the voice input function of the bank application (Application, APP): "I am a sales manager, monthly income 15,000, often fly domestically and internationally, like to sea and dine, and want to apply for a credit card with airport lounge benefits."

[0047] After the intelligent agent (such as an intelligent assistant) receives this item of consultation information, first, a large language model is called to parse the item of consultation information and extract the explicit information and implicit requirements therein, for example:

[0048] The extracted explicit information can include: applying for a credit card, and needing airport lounge benefits.

[0049] The extracted implicit requirements can include: overseas consumption discounts or cashback, accumulation of airline miles, dining discounts, and payment convenience and discounts related to sea shopping, etc.

[0050] According to the explicit information and implicit needs extracted by the large language model, the agent determines that the recommendation process needs to be started for the user. The next step is to extract the user's attribute information, such as user occupation: sales manager, user income: monthly income 15,000, etc.

[0051] At the same time, the user's portrait state is dynamically updated according to the obtained attribute information, for example:

[0052] User name: Zhang XX;

[0053] Age: 32;

[0054] Gender: male;

[0055] Occupation: sales manager;

[0056] Income: monthly income 15,000;

[0057] Consumption habits: frequently fly domestically and internationally, like to sea and dine;

[0058] Credit card needs: airport lounge benefits, overseas consumption discounts, and air mileage accumulation.

[0059] When the user portrait state is updated to meet the recommendation conditions (for example, including occupation, income, consumption habits, and explicit credit card needs), one or more credit cards that match the target portrait corresponding to the current user portrait state are retrieved from the pre-constructed multi-source knowledge base, for example:

[0060] A bank business travel platinum card: provides unlimited airport lounge, overseas consumption cashback, and air mileage accumulation.

[0061] B bank high-end credit card: provides designated airport lounge, dining discount, and sea payment convenience.

[0062] C bank aviation co-branded card: provides rapid air mileage accumulation and ticket discount.

[0063] The multi-source knowledge base stores the latest bank knowledge reserves, such as including the bank's internal product database, market information, etc., which provides detailed information of all recommendable credit cards and is independent of user behavior.

[0064] Then, according to the retrieved credit cards and user portrait, the large language model is used to generate recommendation reasons, for example:

[0065] "Dear Mr. Zhang XX, according to your occupation, income, and consumption habits, we recommend the following credit cards for you:

[0066] A Bank Business Platinum Card: This card is very suitable for you who often travel, providing unlimited access to airport lounges, allowing you to enjoy comfortable rest during your journey. At the same time, overseas consumption cashback and airline mileage accumulation can also save you expenses.

[0067] B Bank High-end Credit Card: This card provides designated airport lounges, as well as dining discounts and convenient payment for overseas shopping, meeting your needs for overseas shopping and dining.

[0068] C Bank Air Co-branded Card: If you pay more attention to the accumulation of airline miles, this card can help you quickly accumulate miles and enjoy discounts on air tickets.

[0069] Finally, the matched credit cards and generated recommendation reasons are presented to the user.

[0070] According to the embodiments of the present application, since the LLM is introduced to deeply mine the user's consultation information, the user's potential needs can be more comprehensively and accurately understood, and on this basis, the user portrait dynamic updating mechanism + multi-source database retrieval + LLM recommendation reason generation, the mutual coordination between such a workflow engine, can adjust the recommendation according to the dynamic changes of the user portrait and multi-dimensional information, improve the system resource utilization, and also ensure the timeliness and comprehensiveness of the recommendation results, thereby comprehensively improving the personalization degree of the recommendation and the ability of the system to process complex information.

[0071] In the embodiments of the present application, before operation S210 responds to receiving the user input consultation information, the consultation information is analyzed by using a large language model to obtain the explicit information and implicit demand in the consultation information, for example, can include:

[0072] In response to the user initiating a consultation request, a pre-configured multi-modal input interface is used to receive the consultation request; wherein the multi-modal input interface is configured to be able to receive multiple modal data in the form of text, voice, image, and document.

[0073] In the embodiments of the present application, through the pre-configured multi-modal input interface, the way the agent receives the user's consultation request is expanded, so that it can process multiple modal data such as text, voice, image and document, as shown in Figure 3 .

[0074] Figure 3 The schematic diagram shows the principle diagram of the multi-modal input interface receiving input data according to the embodiments of the present application.

[0075] As shown in Figure 3 , in the embodiments of the present application, the multi-modal input interface supports multiple modal data input, for example:

[0076] Text input: As the user can input the text "I want to apply for a credit card suitable for frequent business trips" in the chat interface of the APP.

[0077] Voice input: As the user can click the voice input button of the APP and say "I want to apply for a credit card suitable for frequent business trips" into the microphone.

[0078] Image input: As the user takes a photo of a ticket, the APP extracts the user's travel information such as destination, travel frequency, etc. through image recognition technology.

[0079] Document input: As the user uploads a personal financial statement, the APP extracts the user's income, expenditure, etc. through document parsing technology.

[0080] The agent receives these information through the multi-modal input interface and integrates them into a unified consultation request.

[0081] According to the embodiments of the present application, the convenience and flexibility of user interaction with the agent are improved, so that the user can express his needs in the most natural and convenient way, thereby improving the consultation efficiency and user satisfaction.

[0082] In the embodiments of the present application, operation S210 responds to receiving user input consultation information, and uses a large language model to analyze the consultation information to obtain explicit information and implicit requirements in the consultation information, which can include, for example:

[0083] In response to receiving user input consultation information, a large language model is used to identify named entities in the consultation information and perform intent recognition on the named entities to obtain explicit information. According to the named entities, the user's potential requirements are predicted using prompt words to obtain implicit requirements, wherein the prompt words are used to guide the large language model to perform reasoning tasks according to the specified path.

[0084] In the embodiments of the present application, the LLM identifies the named entities in the consultation information and performs intent recognition to obtain the explicit information; at the same time, the prompt words are used to predict the user's potential requirements to obtain the implicit requirements, so as to more comprehensively understand the user's intention.

[0085] For example, when the user inputs in the agent: "I often travel on business and want to apply for a credit card".

[0086] The LLM first performs named entity recognition, such as recognizing the named entity "business trip". Then it performs intent recognition, and the LLM identifies the user's intent as "applying for a credit card". In this way, the explicit information of this time can be obtained: the user needs to apply for a credit card and often travels on business.

[0087] After identifying the explicit information, further according to the prompt word engineering to dig the user's potential requirements.

[0088] For example, based on the named entity "business trip", the prompt "What credit card benefits might frequent business travelers need?" guides the LLM to reason and arrive at the following potential needs:

[0089] Airport VIP lounges, airline mileage redemption, and cashback on overseas spending.

[0090] This allows us to uncover the hidden needs: airport lounges, frequent flyer miles redemption, and cashback on overseas spending.

[0091] According to the embodiments of this application, the accuracy and depth of LLM's understanding of user consultation information are improved, enabling the intelligent agent to better grasp the user's real needs and provide a more reliable foundation for subsequent personalized recommendations.

[0092] In this embodiment of the application, operation S220 extracts the user's attribute information based on explicit information and implicit requirements, which may include, for example:

[0093] Based on explicit information and implicit needs, obtain the user's authorization to extract income information, and extract the user's income information after obtaining the user's authorization; in response to the successful extraction of income information, obtain the user's authorization to extract occupation information, and extract the user's occupation information after obtaining the user's authorization; in response to the successful extraction of occupation information, obtain the user's authorization to extract consumption habits, and extract the user's consumption habits after obtaining the user's authorization.

[0094] In this embodiment of the application, the user's attribute information (income, occupation, consumption habits) is obtained through a step-by-step authorization approach, which fully respects the user's right to privacy and avoids excessive collection of user information.

[0095] For example, in embodiments of this application, user consent or authorization can be obtained before acquiring user attribute information. Specifically, the intelligent agent can employ the following steps when acquiring user attribute information:

[0096] Income Information Authorization: The agent sends a prompt to the user: "In order to better recommend credit cards to you, we need to obtain your income information. Do you want to authorize this?"

[0097] If the user authorizes the process, the agent extracts income information using data such as bank statements and pay slips provided by the user. If the user refuses authorization, the income information extraction step is skipped.

[0098] Occupational Information Authorization: If income information is successfully extracted, the agent sends a prompt to the user: "In order to better recommend credit cards to you, we need to obtain your occupational information. Do you want to authorize it?"

[0099] If the user authorizes, the intelligent agent extracts professional information through the business card, work certificate and other information provided by the user. If the user refuses authorization, the professional information extraction step is skipped.

[0100] Consumption habit authorization: if the professional information extraction is successful, the intelligent agent sends a prompt message to the user: "In order to better recommend credit cards for you, we need to obtain your consumption habits. Do you authorize?".

[0101] If the user authorizes, the intelligent agent extracts consumption habits through the consumption records, bills and other information provided by the user. If the user refuses authorization, the consumption habit extraction step is skipped.

[0102] According to the embodiments of the present application, the user's trust in the intelligent agent is enhanced, the user's willingness to provide information is improved, and more comprehensive and accurate user portrait information can be obtained, thereby providing better support for subsequent personalized recommendation.

[0103] In the embodiments of the present application, the user portrait state is updated according to the attribute information in operation S220, which may include, for example:

[0104] Obtain the user's authorization for extracting identity information, and extract the user's identity information after obtaining the user's authorization for extracting identity information; construct an initial portrait of the user according to the identity information; update the state of the initial portrait based on the initial portrait, the income information, the professional information and the consumption habit, to obtain a target portrait.

[0105] In the embodiments of the present application, the initial portrait of the user is constructed based on the identity information, and the state of the initial portrait is updated using attribute information such as income information, professional information and consumption habit, so as to gradually improve the user portrait.

[0106] In the embodiments of the present application, the user's consent or authorization can be obtained before the user's identity information is obtained.

[0107] For example, after the intelligent agent obtains the user's authorization for extracting identity information, the user's name, age, gender and other identity information are extracted through the user-provided identity card information.

[0108] After extracting the user's name, age, gender and other basic identity information, the following operations can be performed:

[0109] Initial portrait construction: construct an initial portrait of the user according to the identity information, for example:

[0110] Name: Zhang XX;

[0111] Age: 32 years old;

[0112] Gender: male;

[0113] Image state update: based on the initial image, the state of the initial image is updated using income information, occupation information and consumption habits, for example:

[0114] Income: monthly income 15,000;

[0115] Occupation: sales manager;

[0116] Consumption habits: often fly at home and abroad, like to sea and dining;

[0117] Finally, the updated user image is used as the target image of the user.

[0118] According to the embodiments of the present application, the characteristics of the user can be more comprehensively and accurately described, and a more reliable basis is provided for subsequent personalized recommendation.

[0119] In the embodiments of the present application, the updating of the user image state according to the attribute information in operation S220 may, for example, further include:

[0120] In response to the target image not meeting the recommendation condition, the user's authorization for extracting historical behavior data is obtained, and the historical behavior data of the user is extracted after obtaining the user's authorization for extracting historical behavior data; the target image is updated according to the historical behavior data.

[0121] In the embodiments of the present application, when the target image does not meet the recommendation condition, the historical behavior data of the user is further extracted as supplementary information to continue to improve the user image and improve the accuracy of the recommendation.

[0122] For example, the agent will determine whether the current target image meets the recommendation condition, for example: whether it contains sufficient attribute information, whether it can determine the user's demand, etc.

[0123] If the target image does not meet the recommendation condition, the agent will send a prompt message to the user: "In order to better recommend credit cards for you, we need to obtain your historical behavior data, do you authorize?".

[0124] When the user's authorization is obtained, the agent extracts the historical behavior data through the bank flow, consumption records and other information provided by the user.

[0125] Then, the target image is updated according to the historical behavior data, for example:

[0126] The user often consumes abroad, and the amount is high.

[0127] The user often buys airline tickets.

[0128] The updated user image contains more detailed historical behavior data and can more accurately describe the characteristics of the user.

[0129] According to the embodiments of the present application, the personal habits of the user can be understood more deeply, and more accurate basis is provided for subsequent personalized recommendation.

[0130] In the embodiments of the present application, operation S230 retrieves at least one target credit card matched with the target portrait from the pre-constructed multi-source knowledge base according to the target portrait corresponding to the current user portrait state in response to the user portrait state satisfying the recommendation condition, which may, for example, include:

[0131] The matching degree between the target portrait and all credit cards that can be retrieved in the multi-source knowledge base is calculated, the retrieved credit cards are sorted according to the matching degree, and the top-k credit cards with the highest matching degree are selected as the target credit cards based on the sorted credit cards.

[0132] In the embodiments of the present application, the matching degree between the target portrait and all credit cards in the multi-source knowledge base is calculated, and the credit card most suitable for the user is selected according to the top-k filtering mechanism.

[0133] For example, the agent calculates the matching degree between the target portrait and all credit cards in the multi-source knowledge base, assuming that the target portrait contains the feature of “frequent business trips”, and a credit card provides the benefit of “airport lounge”, then the matching degree between the credit card and the target portrait is higher.

[0134] Then, the retrieved credit cards are sorted according to the matching degree, for example, the credit card with the highest matching degree is placed at the front.

[0135] Finally, the top-k credit cards with the highest matching degree are selected as the target credit cards based on the sorted credit cards, for example, the top-3 credit cards with the highest matching degree can be selected as the target credit cards.

[0136] According to the embodiments of the present application, the credit card most matched with the user's demand can be found more accurately, and the accuracy of recommendation and user satisfaction can be improved.

[0137] In the embodiments of the present application, operation S240 generates a recommendation reason using a large language model according to the at least one target credit card and the target portrait, which may, for example, include:

[0138] The information of the at least one target credit card and the information of the target portrait are filled into a preset template, wherein the preset template matches the target portrait; and the large language model is guided to generate a recommendation reason corresponding to the preset template according to the preset template.

[0139] In the embodiments of the present application, the information of the target credit card and the information of the target portrait are filled into the preset template, and the LLM is guided to generate a recommendation reason corresponding to the preset template, so as to ensure the structuring, personalization and quality of the recommendation reason.

[0140] For example, the agent first selects a matching preset template according to the target portrait. Assuming that the user portrait involves "frequent business trips", the "frequent business traveler credit card recommendation template" can be selected.

[0141] Then fill in the information of the target credit card and the information of the target portrait into the preset template, for example:

[0142] Credit card name: XX Bank Air Co-branded Platinum Card.

[0143] Core benefits: airport lounge, airline mileage redemption, overseas consumption cashback.

[0144] User name: Zhang XX.

[0145] User occupation: Sales Manager.

[0146] User consumption habit: frequently fly domestically and internationally.

[0147] Finally, the agent guides the LLM to generate a recommendation reason corresponding to the preset template according to the preset template, for example:

[0148] Dear Mr. Zhang XX, the XX Bank Air Co-branded Platinum Card is perfect for you! As a sales manager who frequently travels domestically and internationally, this card provides airport lounge services to ensure your comfort during travel. In addition, airline mileage redemption and overseas consumption cashback can also save you money when traveling.

[0149] According to the embodiments of the present application, high-quality recommendation reasons can be generated more efficiently, improving the persuasiveness of the recommendation, thereby promoting the user to apply for a credit card.

[0150] Figure 4 The principle diagram of the credit card recommendation method according to the embodiments of the present application is schematically shown.

[0151] As Figure 4 shown in the embodiments of the present application, the principle of the credit card recommendation method is as follows:

[0152] (1) User input consultation: The user can express their credit card needs through natural language or other means, such as "I want to apply for a credit card suitable for frequent business trips", or provide more detailed information about their occupation, income, consumption habits, etc.

[0153] (2) LLM analysis: The LLM analyzes the user's consultation information and extracts explicit and implicit needs. Explicit needs are needs that the user explicitly expresses, such as "need for airport lounge benefits". Implicit needs are needs that the user does not explicitly express but can be inferred from other information about the user, such as inferring the need for overseas consumption discounts or airline mileage accumulation based on "frequent business trips".

[0154] (3) User portrait update: According to the analysis result of LLM, the system updates the user portrait. The user portrait is a description of the user's characteristics, including the user's basic information (such as age, gender, occupation, income), consumption habits, interests, credit card needs, etc. The update of the user portrait is a dynamic process, and the user portrait will be continuously improved and adjusted as the user information is continuously input and the behavior is continuously occurring.

[0155] (4) Recommendation condition judgment: The system judges whether the current user portrait meets the recommendation condition. The recommendation condition can include the completeness of the user portrait, the accuracy of the information, and whether it contains enough information to support the recommendation. If the recommendation condition is not met, it will enter the "supplement information" link.

[0156] (5) Supplement information: The system collects more information through interaction with the user to improve the user portrait. For example, the system can ask the user's income range, consumption behavior, whether there is a specific credit card brand, etc. The collected information will be used to update the user portrait.

[0157] (6) Knowledge base retrieval: When the user portrait meets the recommendation condition, the system retrieves the credit card matching the user portrait from the multi-source knowledge base. The knowledge base contains information about various credit cards, such as credit card benefits, applicable population, etc. The retrieval process combines various dimensions of the user portrait, such as income, occupation, consumption habits, credit card needs, etc., to select the most suitable credit card.

[0158] (7) Generate recommendation reason and recommend: The system generates personalized recommendation reasons based on the retrieved credit card and user portrait. The recommendation reason explains why these credit cards are recommended to the user and highlights the matching between the credit card and the user's needs. For example, "According to your frequent business trip needs, we recommend you XX airline co-branded card, which provides airport lounge services and airline mileage accumulation." Finally, the system presents the credit card and the recommendation reason to the user.

[0159] In addition, user feedback can also be collected, and the user can choose to accept the recommendation or choose not to accept the recommendation and make modifications. If the user is not satisfied with the recommendation result, the system will adjust the user portrait or re-perform knowledge base retrieval according to the user's feedback, and then generate the recommendation reason and present the recommendation result again until the user is satisfied.

[0160] Based on the above credit card recommendation method, the present application also provides a credit card recommendation device. The following will combine Figure 5 to describe the device in detail.

[0161] Figure 5 The structure block diagram of the credit card recommendation device according to the embodiment of the present application is schematically shown.

[0162] As shown in Figure 5 The credit card recommendation device 500 of this embodiment includes an analysis module 510, an extraction module 520, a retrieval module 530, and a recommendation module 540.

[0163] The analysis module 510 is configured to analyze the consultation information using a large language model in response to receiving the consultation information input by the user, and obtain explicit information and implicit requirements in the consultation information. In an embodiment, the analysis module 510 can be configured to perform the operation S210 described above, and details are not repeated here.

[0164] The extraction module 520 is configured to extract attribute information of the user according to the explicit information and the implicit requirements, and update the user portrait state according to the attribute information. In an embodiment, the extraction module 520 can be configured to perform the operation S220 described above, and details are not repeated here.

[0165] The retrieval module 530 is configured to retrieve at least one target credit card matching a target portrait from a pre-constructed multi-source knowledge base in response to the user portrait state satisfying a recommendation condition, the target portrait corresponding to the current user portrait state. In an embodiment, the retrieval module 530 can be configured to perform the operation S230 described above, and details are not repeated here.

[0166] The recommendation module 540 is configured to generate a recommendation reason using a large language model according to the at least one target credit card and the target portrait, and recommend the at least one target credit card and the recommendation reason to the user. In an embodiment, the recommendation module 540 can be configured to perform the operation S240 described above, and details are not repeated here.

[0167] According to an embodiment of the present application, the credit card recommendation device 500 further includes a receiving module.

[0168] The receiving module is configured to receive a consultation request using a pre-configured multi-modal input interface in response to the user initiating the consultation request; wherein the multi-modal input interface is configured to be able to receive multi-modal data in the form of text, voice, image, and document.

[0169] According to an embodiment of the present application, the analysis module 510 includes an identification module and a prediction module.

[0170] The identification module is configured to identify a named entity in the consultation information using a large language model in response to receiving the consultation information input by the user, and perform intent identification on the named entity to obtain explicit information.

[0171] The prediction module is configured to predict potential requirements of the user using a prompt word according to the named entity to obtain implicit requirements, wherein the prompt word is used to guide the large language model to perform an inference task according to a specified path.

[0172] According to an embodiment of the present application, the extraction module 520 comprises: an income information extraction module, a profession information extraction module and a consumption habit extraction module.

[0173] The income information extraction module is configured to obtain authorization of the user for extracting income information according to the explicit information and the implicit demand, and extract the income information of the user after obtaining the authorization of the user for extracting the income information.

[0174] The profession information extraction module is configured to obtain authorization of the user for extracting profession information in response to successful extraction of the income information, and extract the profession information of the user after obtaining the authorization of the user for extracting the profession information.

[0175] The consumption habit extraction module is configured to obtain authorization of the user for extracting consumption habit in response to successful extraction of the profession information, and extract the consumption habit of the user after obtaining the authorization of the user for extracting the consumption habit.

[0176] According to an embodiment of the present application, the extraction module 520 further comprises: an identity information extraction module, a construction module and a first updating module.

[0177] The identity information extraction module is configured to obtain authorization of the user for extracting identity information, and extract the identity information of the user after obtaining the authorization of the user for extracting the identity information.

[0178] The construction module is configured to construct an initial portrait of the user according to the identity information.

[0179] The first updating module is configured to update a state of the initial portrait based on the initial portrait, the income information, the profession information and the consumption habit, and obtain a target portrait.

[0180] According to an embodiment of the present application, the extraction module 520 further comprises: a historical behavior data extraction module and a second updating module.

[0181] The historical behavior data extraction module is configured to obtain authorization of the user for extracting historical behavior data in response to the target portrait not satisfying a recommendation condition, and extract the historical behavior data of the user after obtaining the authorization of the user for extracting the historical behavior data.

[0182] The second updating module is configured to update the target portrait according to the historical behavior data.

[0183] According to an embodiment of the present application, the retrieval module 530 comprises: a calculation module, a sorting module and a screening module.

[0184] The calculation module is configured to calculate a matching degree between the target portrait and all credit cards that can be retrieved in the multi-source knowledge base.

[0185] The sorting module is configured to sort the retrieved credit cards according to the matching degree.

[0186] The screening module is configured to screen the top k credit cards with the highest matching degrees as target credit cards based on the sorted credit cards.

[0187] According to an embodiment of the present application, the recommendation module 540 comprises a filling module and a guiding module.

[0188] The filling module is configured to fill information of at least one target credit card and information of the target portrait into a preset template, wherein the preset template matches the target portrait.

[0189] The guiding module is configured to guide the large language model to generate a recommendation reason corresponding to the preset template according to the preset template.

[0190] According to an embodiment of the present application, any one or more of the parsing module 510, the extraction module 520, the retrieval module 530 and the recommendation module 540 can be combined in one module, or any one of them can be split into multiple modules. Alternatively, at least part of the function of one or more of these modules can be combined with at least part of the function of other modules, and implemented in one module. According to an embodiment of the present application, at least one of the parsing module 510, the extraction module 520, the retrieval module 530 and the recommendation module 540 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging circuits, etc. hardware or firmware, or any one of software, hardware and firmware or any appropriate combination of any of them. Alternatively, at least one of the parsing module 510, the extraction module 520, the retrieval module 530 and the recommendation module 540 can be at least partially implemented as a computer program module which can perform corresponding functions when running.

[0191] Figure 6 The block diagram schematically shows an electronic device suitable for implementing the credit card recommendation method according to an embodiment of the present application.

[0192] As Figure 6As shown, the electronic device 600 according to embodiments of the present application includes a processor 601 which can perform various appropriate actions and processes in accordance with a program stored in a read only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The processor 601 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chip set, and / or a special purpose microprocessor (e.g., an application specific integrated circuit (ASIC)), and so on. The processor 601 can also include an on-board memory for cache use. The processor 601 can include a single processing unit or multiple processing units for executing different actions of the method processes according to embodiments of the present application.

[0193] In the RAM 603, various programs and data required for the operation of the electronic device 600 are stored. The processor 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. The processor 601 performs various operations of the method processes according to embodiments of the present application by executing the programs in the ROM 602 and / or the RAM 603. Note that the programs can also be stored in one or more memories other than the ROM 602 and the RAM 603. The processor 601 can also perform various operations of the method processes according to embodiments of the present application by executing the programs stored in the one or more memories.

[0194] According to embodiments of the present application, the electronic device 600 can also include an input / output (I / O) interface 605 which is also connected to the bus 604. The electronic device 600 can also include one or more of the following components connected to the input / output (I / O) interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as necessary. A removable medium 611 such as a magnetic disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 610 as necessary, so that a computer program read out therefrom is installed in the storage section 608 as necessary.

[0195] The application further provides a computer readable storage medium, which can be included in the device / apparatus / system described in the above embodiments, or can exist independently without being assembled into the device / apparatus / system. The computer readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the application.

[0196] According to the embodiments of the application, the computer readable storage medium can be a non-volatile computer readable storage medium, which can include, but is not limited to, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any appropriate combination thereof. In this application, a computer readable storage medium can be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to the embodiments of the application, the computer readable storage medium can include one or more of the above-described ROM 602 and / or RAM 603 and / or one or more memories other than the ROM 602 and the RAM 603.

[0197] The embodiments of the application also include a computer program product, which includes a computer program containing program codes for executing the methods shown in the flowcharts. When the computer program product is run in a computer system, the program codes are used to make the computer system implement the methods provided by the embodiments of the application.

[0198] The above-described functions defined in the system / apparatus of the embodiments of the application are performed when the computer program is executed by the processor 601. According to the embodiments of the application, the above-described system, apparatus, module, unit, etc. can be implemented by computer program modules.

[0199] In one embodiment, the computer program can rely on a tangible storage medium such as an optical storage medium, a magnetic storage medium, etc. In another embodiment, the computer program can also be transmitted, distributed, and downloaded in the form of a signal on a network medium, and be downloaded and installed through the communication part 609, and / or installed from the detachable medium 611. The program codes contained in the computer program can be transmitted by any appropriate network medium, including but not limited to wireless, wired, etc., or any appropriate combination thereof.

[0200] In such embodiments, the computer program can be downloaded and installed from the network via the communication section 609, and / or installed from the removable media 611. When the computer program is executed by the processor 601, the above-described functions defined in the system of the embodiments of the present application are executed. According to the embodiments of the present application, the system, device, apparatus, module, unit, and the like described above can be realized by the computer program modules.

[0201] According to the embodiments of the present application, the program code for executing the computer program provided by the embodiments of the present application can be written in any combination of one or more programming languages, and specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming language, and / or assembly / machine language. The programming language includes, but is not limited to, such as Java, C++, python, "C" language, or similar programming language. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case involving a remote computing device, the remote computing device can be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, connected to the Internet through an Internet service provider).

[0202] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectural, functional, and operational scenarios of systems, methods, and computer program products in accordance with various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment, or a portion of code, which includes one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in a different order than that noted in the accompanying diagrams. For example, two blocks noted in succession can actually be executed substantially concurrently or in the opposite order, depending on the functionality involved. It should also be noted that each block in the flowcharts or block diagrams, and combinations of blocks in the flowcharts or block diagrams, can be implemented by dedicated hardware-based systems that perform the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0203] Those skilled in the art can understand that the features described in various embodiments of the present application can be combined and / or integrated in various combinations and / or integrations, even if such combinations or integrations are not explicitly described in the present application. In particular, the features described in various embodiments of the present application can be combined and / or integrated in various combinations and / or integrations without departing from the spirit and teachings of the present application. All such combinations and / or integrations fall within the scope of the present application.

Claims

1. A credit card recommendation method characterized by, The method applied to an intelligent agent comprises: In response to receiving user input consultation information, analyzing the consultation information using a large language model to obtain explicit information and hidden demand in the consultation information; According to the explicit information and the hidden demand, extracting attribute information of the user, and updating the user portrait state according to the attribute information; In response to the user portrait state satisfying the recommendation condition, retrieving at least one target credit card matched with the target portrait from a pre-constructed multi-source knowledge base according to the target portrait corresponding to the current user portrait state; According to the at least one target credit card and the target portrait, generating a recommendation reason using the large language model, and recommending the at least one target credit card and the recommendation reason to the user.

2. The method of claim 1, wherein, Before the response to receiving user input consultation information, using a large language model to analyze the consultation information, obtaining explicit information and hidden demand in the consultation information, comprising: In response to the user initiating a consultation request, receiving the consultation request using a pre-configured multi-modal input interface; Wherein, the multi-modal input interface is configured to be able to receive multiple modal data in the form of text, voice, image and document.

3. The method of claim 2, wherein, The response to receiving user input consultation information, using a large language model to analyze the consultation information, obtaining explicit information and hidden demand in the consultation information, comprising: In response to receiving user input consultation information, using a large language model to identify named entities in the consultation information, and performing intent recognition on the named entities to obtain explicit information; According to the named entity, using prompt word to predict the potential demand of the user to obtain hidden demand, wherein the prompt word is used to guide the large language model to perform reasoning task according to the specified path.

4. The method of claim 1, wherein, According to the explicit information and the hidden demand, extracting attribute information of the user, comprising: According to the explicit information and the hidden demand, obtaining the user's authorization to extract income information, and extracting the user's income information after obtaining the user's authorization to extract income information; In response to the successful extraction of the income information, obtaining the user's authorization to extract occupation information, and extracting the user's occupation information after obtaining the user's authorization to extract occupation information; In response to the successful extraction of the occupation information, obtaining the user's authorization to extract consumption habits, and extracting the user's consumption habits after obtaining the user's authorization to extract consumption habits.

5. The method of claim 4, wherein, The response to receiving user input consultation information, using a large language model to analyze the consultation information, obtaining explicit information and hidden demand in the consultation information, comprising: Obtaining the user's authorization to extract identity information, and extracting the user's identity information after obtaining the user's authorization to extract identity information; According to the identity information, constructing an initial portrait of the user; Based on the initial portrait, updating the state of the initial portrait using the income information, the occupation information and the consumption habits to obtain a target portrait.

6. The method of claim 5, wherein, The response to receiving user input consultation information, using a large language model to analyze the consultation information, obtaining explicit information and hidden demand in the consultation information, further comprising: In response to the target portrait not satisfying the recommendation condition, obtaining the user's authorization to extract historical behavior data, and extracting the user's historical behavior data after obtaining the user's authorization to extract historical behavior data; According to the historical behavior data, updating the target portrait.

7. The method of claim 1, wherein, The method comprises the following steps: calculating the matching degree between the target portrait and all credit cards that can be retrieved in the multi-source knowledge base; sorting the retrieved credit cards according to the matching degree; based on the sorted credit cards, filtering out the top k credit cards with the highest matching degree as target credit cards.

8. The method of claim 1, wherein, The method comprises the following steps: filling the information of the at least one target credit card and the information of the target portrait into a preset template, wherein the preset template matches the target portrait; according to the preset template, guiding the large language model to generate a recommendation reason corresponding to the preset template.

9. A credit card recommendation device characterized by comprising: The device comprises: an analysis module configured to analyze the consultation information using a large language model in response to receiving user inputted consultation information, and obtain explicit information and implicit requirements in the consultation information; an extraction module configured to extract attribute information of a user according to the explicit information and the implicit requirements, and update a user portrait state according to the attribute information; a retrieval module configured to retrieve at least one target credit card matching a target portrait from a pre-constructed multi-source knowledge base in response to the user portrait state meeting a recommendation condition; a recommendation module configured to generate a recommendation reason using the large language model according to the at least one target credit card and the target portrait, and recommend the at least one target credit card and the recommendation reason to the user.

10. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1-8.

11. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed by the processor to implement the steps of the method according to any one of claims 1-8.

12. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions are executed by the processor to implement the steps of the method according to any one of claims 1-8.

Images