Credit investigation service recommendation system, method, device, storage medium and program product

Abstract

The one or more embodiments of the specification provide a credit investigation service recommendation system, method, device, storage medium and program product, the system comprising a feature mining engine and an intelligent recommendation engine, wherein the feature mining engine generates a prompt word according to a target institution portrait of a target institution and a credit investigation service portrait of each credit investigation service to be recommended and inputs a large language model to obtain a target correlation between the target institution portrait and each credit investigation service portrait output by the model; and generates a prompt word according to the corresponding correlation data of the target institution and inputs a large language model to obtain the intention feature of the target institution output by the model. The intelligent recommendation engine determines the target credit investigation service meeting the final demand of the institution from each credit investigation service to be recommended according to the corresponding correlation of the institution and the intention feature.

Description

TECHNICAL FIELD

[0001] One or more embodiments of the present specification relate to the technical field of credit service recommendation, and in particular to a credit service recommendation system, method, device, storage medium and program product. BACKGROUND

[0002] Credit service recommendation refers to intelligently predicting and recommending credit services that may be of interest to an institution through comprehensive data analysis. However, in the current related technology, credit service recommendation relies on artificial rules or simple label matching, which is difficult to cope with complex and variable credit service scenarios. For example, with the continuous improvement of China's credit system, the service range of market-oriented credit institutions continues to expand, and the types of credit services are also increasingly rich. The current credit services can cover identity authentication, fraud identification, credit risk assessment, debt repayment ability analysis and other categories, and can meet the credit service use needs of banks, consumer companies, resource lending companies, guarantee institutions and other institutions.

[0003] Therefore, how to improve the accuracy of credit service recommendation to meet complex and variable credit service scenarios is a technical problem to be solved at present. SUMMARY

[0004] Therefore, one or more embodiments of the present specification provide a credit service recommendation system, method, device, storage medium and program product.

[0005] To achieve the above-mentioned purpose, one or more embodiments of the present specification provide technical solutions as follows:

[0006] According to a first aspect of one or more embodiments of the present specification, a credit service recommendation system is provided, comprising:

[0007] a feature mining engine configured to generate a prompt word according to a target institution portrait of a target institution and a credit service portrait of each credit service to be recommended and input a large language model to obtain a target association relationship between the target institution portrait and each credit service portrait output by the model; and generate a prompt word according to the association data corresponding to the target institution and input the large language model to obtain the intention feature of the target institution output by the model; the association data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution;

[0008] an intelligent recommendation engine configured to determine each candidate credit service matched with the target institution from the each credit service to be recommended according to the target association relationship, and determine a target credit service conforming to the intention of the target institution from the each candidate credit service according to the intention feature, so as to recommend the target credit service to the target institution.

[0009] According to a second aspect of one or more embodiments of the present specification, a credit service recommendation method is provided, the method is applied to an intelligent recommendation engine, and the method comprises:

[0010] In response to a credit service recommendation demand issued by a target institution to a credit service recommendation system, determining each candidate credit service matched with the target institution from each credit service to be recommended according to a target correlation corresponding to the target institution; the target correlation is a correlation between the target institution portrait and each credit service portrait obtained by the feature mining engine by generating a prompt word according to the target institution portrait of the target institution and the credit service portrait of each credit service to be recommended and inputting a large language model;

[0011] According to the intention feature of the target institution, determine the target credit service that meets the intention of the target institution from the candidate credit services, and recommend the target credit service to the target institution; the intention feature is obtained by the feature mining engine by generating a prompt word according to the correlation data corresponding to the target institution and inputting a large language model; the correlation data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution.

[0012] According to a third aspect of one or more embodiments of the present specification, an electronic device is provided, comprising:

[0013] a processor;

[0014] a memory for storing processor-executable instructions;

[0015] The processor implements the method of the second aspect by running the executable instructions.

[0016] According to a fourth aspect of one or more embodiments of the present specification, a computer readable storage medium is provided, which stores computer instructions, and the instructions are executed by a processor to implement the steps of the method of the second aspect.

[0017] According to a fifth aspect of one or more embodiments of the present specification, a computer program product is provided, comprising: computer program / instructions, which are executed by a processor to implement the method of the second aspect.

[0018] As can be known from the above embodiments, the credit service recommendation system, method, device, storage medium and program product provided by one or more embodiments of the present specification have the following advantages. The feature mining engine generates a prompt word according to the target institution portrait of the target institution and the credit service portrait of each credit service to be recommended and inputs the large language model to obtain the target association relationship between the target institution portrait and each credit service portrait output by the model; and generates a prompt word according to the association data corresponding to the target institution and inputs the large language model to obtain the intention feature of the target institution output by the model. The feature mining engine can accurately mine the association relationship between the institution portrait and the credit service portrait, thereby obtaining the preferences of different institutions for credit services. Meanwhile, the feature mining engine can also determine the real intention of the institution according to the association data corresponding to the institution, thereby providing a basis for the intelligent recommendation engine to accurately recommend the target credit service. The intelligent recommendation engine determines the target credit service meeting the final demand of the institution from each credit service to be recommended according to the association relationship and the intention feature corresponding to the institution. Since the preferences of the institution for each credit service and the real demand of the institution are fully considered, the intelligent recommendation engine improves the accuracy of recommending the target credit service for the institution. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 FIG. 1 is a schematic diagram of an architecture of a credit service recommendation application scenario provided by an example embodiment.

[0020] Figure 2 FIG. 2 is a structural schematic diagram of a credit service recommendation system provided by an example embodiment.

[0021] Figure 3 FIG. 3 is a structural schematic diagram of another credit service recommendation system provided by an example embodiment.

[0022] Figure 4 FIG. 4 is a flowchart of constructing an institution portrait provided by an example embodiment.

[0023] Figure 5 FIG. 5 is a flowchart of constructing a credit service portrait provided by an example embodiment.

[0024] Figure 6 FIG. 6 is a flowchart of a credit product recommendation provided by an example embodiment.

[0025] Figure 7 FIG. 7 is a flowchart of a credit service recommendation method provided by an example embodiment.

[0026] Figure 8 FIG. 8 is a structural schematic diagram of a device provided by an example embodiment. DETAILED DESCRIPTION

[0027] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all. According to the embodiments in the specification, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the specification.

[0028] The institutional information (including but not limited to institutional equipment information, institutional personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the specification are information and data authorized by the institution or fully authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation portal for the institution to choose authorization or refusal.

[0029] As described in the background, in the current related art, credit investigation service recommendation relies on artificial rules or simple label matching, which is difficult to cope with complex and variable credit investigation service recommendation scenarios. In particular, with the continuous improvement of China's credit investigation system, the service range of market-oriented credit investigation institutions continues to expand, and the types of credit investigation services are also increasingly rich. The current credit investigation services can cover multiple categories such as identity authentication, fraud identification, credit risk assessment, debt repayment ability analysis, etc., and can meet the credit investigation service use needs of banks, consumer companies, resource lending companies, guarantee institutions and other institutions. Therefore, simply relying on artificial rules or label matching cannot accurately recommend credit investigation services needed by institutions among the various types of credit investigation services.

[0030] To sum up, the embodiment of the present specification proposes a credit service recommendation system, which comprises a feature mining engine and an intelligent recommendation engine. The feature mining engine generates prompt words according to a target institution portrait of a target institution and a credit service portrait of each credit service to be recommended and inputs the prompt words into a large language model to obtain a target correlation between the target institution portrait and each credit service portrait output by the model; and generates prompt words according to the associated data corresponding to the target institution and inputs the prompt words into the large language model to obtain the intention features of the target institution output by the model. The feature mining engine can accurately mine the correlation between the institution portrait and the credit service portrait, thereby obtaining the preferences of different institutions for credit services. At the same time, the feature mining engine can also determine the real intention of the institution according to the associated data corresponding to the institution, thereby providing a basis for the intelligent recommendation engine to accurately recommend target credit services. The intelligent recommendation engine determines the target credit service meeting the final demand of the institution from each credit service to be recommended according to the associated data corresponding to the institution and the intention features. Since the preferences of the institution for each credit service and the real demand of the institution are fully considered, the intelligent recommendation engine improves the accuracy of recommending target credit services for the institution. In addition, the credit service recommendation system provided by the embodiment of the present specification proposes a double portrait system of institution portrait + credit service portrait, and constructs the correlation between the institution and the credit service according to the double portrait system, which becomes the basis for accurate matching of credit services and supports high-dimensional heterogeneous data fusion. Moreover, according to the feature automatic mining mechanism of the large language model, end-to-end conversion from unstructured demand text to computable features is realized.

[0031] Figure 1 FIG. 1 is an architecture schematic diagram of a credit service recommendation application scenario provided by an exemplary embodiment. As shown in the figure, the application scenario can comprise a server 11, a network 12, and a plurality of electronic devices such as a PC (Personal Computer) 13, a mobile phone 14, etc. Figure 1

[0032] The server 11 can be a physical server comprising a standalone host, or the server 11 can be a virtual server carried by a host cluster. In the running process, the server 11 can run the server side program of an application to realize the related functions of the application, such as when the server 11 runs the credit service recommendation program, the credit service recommendation system for providing credit service recommendation can be realized.

[0033] ​The PC 13 and the mobile phone 14 are only some types of electronic devices that can be used by the institution. In fact, the institution can obviously also use electronic devices of other types, such as tablet devices, notebook computers, personal digital assistants (PDAs), wearable devices (such as smart glasses, smart watches, etc.), and the like, and the one or more embodiments of the present specification do not limit this. During operation, the electronic device can run a client-side program of an application to implement the relevant functions of the application, for example, the institution can log in to a credit service recommendation program through the electronic device to serve as a client of the credit service recommendation system described above, and after receiving a recommendation request of the institution at the client, the recommendation request can be sent to the server, and a recommendation result can be returned according to the credit service recommendation system in the server. The application program of the client of the program service can be started and run on the electronic device. The client-side program can be a native application installed on the electronic device, or the client-side program can be an applet, a fast application, or other similar forms. Of course, when using web page technologies such as HTML5 or the like, the relevant functions can be implemented through a page displayed by a browser, and the browser can be a standalone browser application or a browser module embedded in some application.

[0034] The network 12 between the electronic devices such as the PC 13 and the mobile phone 14 and the server 11 can be a wired or wireless network according to the communication mode supported by the corresponding electronic device, and the present specification does not limit this. For example, the PC 13 can support wired and wireless communication at the same time, and wired or wireless communication can be used as needed, while the mobile phone 14 usually only supports wireless communication, and wireless communication can be used.

[0035] Reference Figure 2 A flowchart of a credit service recommendation system provided in the present specification, the credit service recommendation system comprising:

[0036] The feature mining engine 202 is configured to generate a prompt word and input a large language model according to a target institution portrait of a target institution and a credit service portrait of each credit service to be recommended, to obtain a target association relationship between the target institution portrait and each credit service portrait output by the model; and generate a prompt word and input a large language model according to the association data corresponding to the target institution, to obtain an intention feature of the target institution output by the model; the association data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution;

[0037] The intelligent recommendation engine 204 is configured to determine, from the various to-be-recommended credit services, a target credit service that matches the target institution according to the target association relationship, and determine, from the various to-be-recommended credit services, a target credit service that meets the target institution's intention according to the intention feature, so as to recommend the target credit service to the target institution.

[0038] After obtaining the target institution portrait of the target institution and the credit service portraits of the various to-be-recommended credit services, the feature mining engine can first generate a prompt word according to the target institution portrait and the various credit service portraits. The prompt word is used to convey the generation instruction of the association relationship, the reference basis for generating the association relationship, and the format requirement of the generation result to the large language model. For example, a prompt word generated according to the target institution portrait and the various credit service portraits can be: please analyze and mine the implicit rules that the target institution portrait prefers a certain type of credit service portrait according to the target institution portrait and the various credit service portraits, and generate an expression representing the target association relationship between the target institution portrait and the various credit service portraits in the format of XXXXX. It should be noted that in the embodiments of the present specification, the target institution can refer to an institution that currently submits a recommendation request, or any institution that has checked into the credit service recommendation system, which is not limited.

[0039] In order to accurately obtain the real intention of the target institution, the feature mining engine generates a prompt word according to the association data corresponding to the target institution and inputs the large language model to obtain the intention feature of the target institution output by the model. Through the large language model, the key intention of the institution can be recognized from the unstructured association data, so as to analyze the real intention of the target institution, so as to better meet the real demand of the target institution. The association data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution. That is, the association data includes only the historical data generated by the target institution in the credit service recommendation system, or the association data includes only the current demand data input by the target institution, or the association data includes both the historical data generated by the target institution in the credit service recommendation system and the current demand data input by the target institution. In some embodiments, the historical data generated by the target institution in the credit service recommendation system includes: the calling log of the target institution in the credit service system, the BD communication record, the credit service information configured by the target institution in the system, and the institution feedback information, etc. In some embodiments, the current demand data can be text demand information input by the target institution when the target institution issues a credit service recommendation demand to the credit service recommendation system. For example, a certain demand text information can be "find a credit service that can identify false income certificates".

[0040] The intelligent recommendation engine can first obtain the target association relationship corresponding to the target institution and the intention feature of the target institution from the feature mining engine when recommending the target credit investigation service to the target institution, and then determine the candidate credit investigation service matched with the target institution from the various credit investigation services to be recommended according to the target association relationship. Optionally, the candidate credit investigation service matched with the target institution can be multiple. After determining the candidate credit investigation service, the target credit investigation service meeting the intention of the target institution is determined from the various candidate credit investigation services according to the intention feature of the target institution, and the target credit investigation service is recommended to the target institution as a recommendation result. It should be noted that the target credit investigation service can be one or more, which is not limited. According to the intention feature, the target credit investigation service meeting the intention of the target institution is determined from the various candidate credit investigation services, which is based on finding the most suitable credit investigation service for the target institution, and combines the intention of the target institution, thereby further improving the accuracy of recommending the credit investigation service to the target institution.

[0041] In some embodiments, the above association relationship can be represented by the association degree between the target institution portrait and each credit investigation service portrait. The association degree can be a value between 0 and 1, and the higher the value, the stronger the association degree. When determining the candidate credit investigation service matched with the target institution from the various credit investigation services to be recommended, the pre-set association degree can be used to filter out the candidate association degree greater than the pre-set association degree from the multiple association degrees, and the credit investigation service corresponding to each candidate association degree is determined as the candidate credit investigation service.

[0042] In some embodiments, the above association relationship can also be represented by the association feature recognized by the model. Then, the association feature of the target institution is input into the model, and the model determines the candidate credit investigation service matched with the target institution from the various credit investigation services to be recommended. It should be noted that the model for determining the candidate credit investigation service can be a large language model, or a traditional model trained by sample data, which is not limited.

[0043] In some embodiments, the various credit investigation services in the credit investigation service recommendation system include: personal credit score, anti-fraud interface, multi-head loan query, identity verification, user risk report, business information check, association graph, etc.

[0044] In some embodiments of the present specification, the feature mining engine is further used to:

[0045] obtain the institutions corresponding to each institution portrait in a plurality of pre-generated institution portraits;

[0046] In response to determining that the target institution is not included in the institutions corresponding to the respective institution portraits, determining a replaceable institution matching the target institution from the institutions corresponding to the respective institution portraits according to attribute information of the target institution;

[0047] determining an institution portrait corresponding to the replaceable institution as the target institution portrait of the target institution.

[0048] It should be noted that the target institution can be an old user who has already checked into the credit service recommendation system, or a new user who has just checked into the credit service recommendation system. The feature mining engine can generate institution portraits in advance according to the relevant data of each institution that has checked into the credit service recommendation system, obtaining a plurality of pre-generated institution portraits. In this way, when an institution portrait of an old user is needed, it can be directly obtained from the pre-generated institution portraits. When the target institution is a new user, i.e., the plurality of pre-generated institution portraits do not directly include the institution portrait of the target institution, in order to improve the efficiency of obtaining the target institution portrait of the target institution, a replaceable institution matching the target institution can be first determined from the institutions corresponding to the respective institution portraits. Specifically, the replaceable institution can be determined from the institutions corresponding to the respective institution portraits according to attribute information of the target institution, which can be information such as type, basic information, business characteristics, compliance records, institution size, risk labels, and other information that can reflect the characteristics of the target institution. The replaceable institution is the institution in the plurality of institutions that is closest to the attribute information of the target institution.

[0049] In some embodiments, the feature mining engine can directly generate an institution portrait according to the relevant data of the target institution when determining the target institution portrait of the target institution, without limitation.

[0050] In some embodiments of the present specification, with reference to Figure 3 , the credit service recommendation system further includes an institution portrait engine 210, configured to:

[0051] For each institution for which an institution portrait is generated, obtain original data information of the institution and new event information related to the institution;

[0052] perform semantic analysis on the original data information and the new event information by a large language model, and obtain first basic feature information of the institution according to a first result of the semantic analysis;

[0053] perform zero-shot classification on the institution according to the original data information and the new event information, and obtain a first classification result of the institution;

[0054] obtain first professional knowledge corresponding to the institution, and obtain first reasoning feature information of the institution according to the first professional knowledge, the original data information and the new event information;

[0055] determine the institution portrait of the institution according to the first basic feature information, the first classification result and the first reasoning feature information.

[0056] In order to accurately generate the institution portrait of the institution, in the embodiment, the credit service recommendation system obtains the original data information of the institution and the new event information related to the institution through the institution portrait engine. The original data information of the institution can be some original data provided by the institution when it enters the credit service system. In some embodiments, the original data information includes the business license, the institution charter, the contract text, the call log, the compliance report, the operation work order and other data information of the institution. The new event information of the institution can be the industrial and commercial change information, the judgment document, the administrative punishment and the public opinion news information of the institution obtained through external data. After obtaining the original data information and the new event information of the institution, the original data information and the new event information can be input into the large language model for semantic analysis, and the first basic feature information of the institution can be obtained according to the first result of semantic analysis. The first basic feature information mainly includes the name, the type, the registered capital, the business scope, the main product, the service group, the daily average call volume, the data use purpose and other basic features of the institution. In some embodiments, the original data information and the new event information can be input into the large language model at the same time, and the prompt word for semantic analysis can also be input. The format of the output result of the large language model can be added in the prompt word, so that the first result output by the large language model can be directly used as the first basic feature information of the institution. For example, the prompt word for semantic analysis can be: "please perform semantic analysis according to the input original data information and new event information, and generate the result in the order of name, type, registered capital, business scope, main product, service group, daily average call volume and data use purpose."

[0057] In the zero-shot classification of the institution based on the original data information and the new event information, the zero-shot classification can be completed by a model. The zero-shot classification refers to the model classifying the current input data into the correct category only through task description and category definition without seeing the training sample corresponding to the current input data. Optionally, the model can be a traditional neural network model or a large language model, which is not limited. When the model is a large language model, the candidate label list and natural language instructions need to be input at the same time as the original data information and the new event information of the institution are input into the large language model. For example, when the original data information and the new event information of an institution are input into a large language model, the following can be input: "Please judge whether the calling behavior of the institution is abnormal, and output the result normal or abnormal". Thus, the large language model completes the zero-shot classification of whether the calling behavior is abnormal. In some embodiments, the candidate label list can be set as needed, for example, a candidate label list can include: whether the institution is likely to use data for collection (corresponding result is or not), whether the institution is often accused of abusing data in user complaints (corresponding result high / medium / low), whether the institution has negative news (corresponding result is or not), whether the institution pays on time (corresponding result is or not), and the like.

[0058] In order to further infer the original data information and the new event information of the institution, and increase the forward-looking of the institution portrait in line with professional knowledge, when the institution is portraited, the first professional knowledge base corresponding to the institution is also acquired. The first professional knowledge base can be a knowledge base related to the institution and having professional guidance significance. The specific first professional knowledge base can be determined by industry regulations, management methods, and administrative regulations related to the institution. For example, in some embodiments, the first professional knowledge base can be determined by the "Credit Reference Management Method". After acquiring the first professional knowledge base, the original data information and the new event information of the institution can be matched and compared with the professional knowledge in the first professional knowledge base, so as to obtain the first inference feature information enhanced by knowledge. In some embodiments, after obtaining the first professional knowledge base, the first professional knowledge base, the original data information and the new event information of the institution are input into a large language model, and the prompt word "mining inference features associated with the first professional knowledge base from the original data information and the new event information of the institution" is input. For example, a first professional knowledge base can be the "Credit Reference Management Method", and the original data information and the new event information of the institution can be that the actual controller of the institution is Zhang, who once served as the CEO of A Company, and A Company was criminally investigated for illegally obtaining citizen information. Through knowledge base matching, Article X of the "Credit Reference Management Method" states that associated personnel with significant illegal records need to be audited. At this time, the first inference feature information that can be output can be that the associated risk level of the institution is high risk.

[0059] After obtaining the first basic feature information, the first classification result and the first inference feature information corresponding to the institution, the above three kinds of information are combined together, and the institution portrait of the institution can be obtained.

[0060] In some embodiments of the present specification, when the institution portrait engine monitors that the change of the new event information of the institution occurs, the institution portrait of the institution is updated according to the changed new event information. For example, after a certain institution generates new business change information, the institution portrait of the institution can be updated according to the new business change information. It should be noted that when the institution portrait is updated by the changed new event information, the first basic feature information, the first classification result and the first inference feature information can be obtained again by combining the changed new event information with the original data information, or the difference information between the changed new event information and the original new event information can be compared, and the corresponding information of the previously generated institution portrait is supplemented or replaced by using the difference information, so as to complete the update of the institution portrait.

[0061] Reference Figure 4 A flowchart for generating an institution portrait provided by an embodiment of the present specification is shown in Figure 4 It can be seen that after obtaining the original data information and the new event information of the institution, the large language model can use the original data information and the new event information to respectively perform semantic extraction, zero-shot classification and knowledge base reasoning to obtain the first basic feature information, the first classification result and the first inference feature information. Among them, when performing knowledge base reasoning, the first professional knowledge base corresponding to the institution needs to be obtained. Finally, according to the first basic feature information, the first classification result and the first inference feature information output by the large language model, the final institution portrait is obtained.

[0062] In some embodiments of the present specification, referring to Figure 3 The credit service recommendation system further includes a credit service portrait engine 212, configured to:

[0063] For each of the plurality of credit services to be recommended, obtain original material information of the credit service to be recommended and change event information related to the credit service to be recommended;

[0064] Perform semantic analysis on the original material information and the change event information by using a large language model, and obtain second basic feature information of the credit service to be recommended according to a second result of the semantic analysis;

[0065] Perform zero-shot classification on the credit service to be recommended according to the original material information and the change event information, and obtain a second classification result of the credit service to be recommended;

[0066] obtain second professional knowledge base corresponding to the to-be-recommended credit service, and obtain second reasoning feature information of the to-be-recommended credit service according to the second professional knowledge base, the original material information, and the change event information;

[0067] obtain a credit service portrait of the to-be-recommended credit service according to the second basic feature information, the second classification result, and the second reasoning feature information.

[0068] In order to accurately generate the institution portrait of the institution, in the embodiment, the credit service recommendation system first obtains the original material information and the change event information of the to-be-recommended credit service through the credit service portrait engine. The original material information can be the original material information provided by the credit service when the credit service is online on the credit service recommendation system. For example, in some embodiments, the original material information can include product specification, product introduction, contract terms, technical interface document, and other material information of the credit service. The change event information can be the change information related to the credit service obtained after the credit service is online. For example, in some embodiments, the change event information can include version update information, user feedback work order, review meeting minutes in the system, and detection compliance statement file of the credit service. After obtaining the original material information and the change event information of the credit service, the original material information and the change event information can be input into the large language model, the second result of the semantic analysis is obtained by the large language model, and the second basic feature information of the institution is obtained according to the second result of the semantic analysis. The second basic feature information mainly includes the credit service name, function description, input field, output field, data source, product type, use scenario, and other information of the credit service. The product type can include identity authentication, fraud identification, credit scoring, debt repayment ability, and the like. In some embodiments, in order to provide the output efficiency of the large language model, the prompt word for semantic analysis can be input at the same time as the original material information and the change event information are input into the large language model, and the format of the output result of the large language model is added in the prompt word, so that the second result output by the large language model can be directly used as the first basic feature information of the credit service. For example, the prompt word for semantic analysis can be: "Please perform semantic analysis according to the input original material information and change event information, and generate the results in the order of credit service name, function description, input field, output field, data source, product type, and use scenario." In some embodiments, the above-mentioned second result can also be converted into the second basic feature information through regular matching.

[0069] In the zero-shot classification of the recommended credit service according to the original material information and the change event information, the zero-shot classification can be completed by a model. Optionally, the model can be a traditional neural network model or a large language model, which is not limited. When the model is a large language model, the candidate label list and the natural language instruction need to be input at the same time as the original material information and the change event information are input into the large language model. For example, when the original material information and the change event information of a certain credit service are input into the large language model, it can be input at the same time whether the credit service involves biological identification information, and the output result is yes or no. Thus, the zero-shot classification of whether the credit service involves biological identification information is completed by the large language model. In some embodiments, the candidate label list can be set as needed, for example, a certain candidate label list can include whether the credit service needs user's separate authorization (corresponding to the result yes or no), which institution types the credit service is suitable for (corresponding to the result bank / consumer institution / small loan company / not applicable), whether the credit service is often complained of misjudgment in user feedback (corresponding to the result positive / neutral / negative), etc.

[0070] In order to further infer the original material information and the change event information in depth and improve the depth and compliance value of the credit service portrait, when the recommended credit service portrait is obtained, a second professional knowledge base of the credit service is also obtained. The second professional knowledge base can be a knowledge base related to the credit service and having professional guidance significance. Specifically, the first professional knowledge base can be determined by industry regulations, management methods, and administrative regulations related to the credit service. For example, in some embodiments, the second professional knowledge base can be determined by the Personal Information Protection Law. After obtaining the second professional knowledge base, the original material information and the change event information of the credit service can be matched and compared with the professional knowledge in the second professional knowledge base, so as to obtain the second inference feature information enhanced by knowledge. In some embodiments, after obtaining the second professional knowledge base, the second professional knowledge base, the original material information and the change event information of the credit service are input into the large language model, and the prompt word is input: "extract the inference features associated with the second professional knowledge base from the original material information and the change event information of the credit service". For example, a certain second professional knowledge base can be the Personal Information Protection Law, and the original material information and the change event information of the credit service can be: the product needs to use face recognition for identity verification, Article X of the Personal Information Protection Law: separate authorization is required for the use of biological identification. At this time, the output first inference feature information can be that the compliance risk level of the credit service is high risk.

[0071] After obtaining the second basic feature information, the second classification result and the second reasoning feature information corresponding to the recommended credit service, the above three kinds of information are combined together to obtain the credit service portrait of the recommended credit service.

[0072] In some embodiments of the present specification, when the credit service portrait engine monitors the change of the change event information of the recommended credit service, it updates the credit service portrait of the recommended credit service according to the changed change event information. For example, after listening to the product change log of a certain credit service, the credit service portrait can be updated according to the product change log. It should be noted that when updating the credit service portrait through the changed change event information, the second basic feature information, the second classification result and the second reasoning feature information can be obtained again by combining the changed change event information with the original material information, or the difference information between the changed change event information and the changed change event information can be compared, and the corresponding information of the previously generated credit service portrait can be supplemented or replaced to complete the update of the credit service portrait.

[0073] In some embodiments, in order to improve the accuracy of zero-shot classification by a large language model, one or more preset examples can be input into the large language model before zero-shot classification of the institution by the large language model, to help the large language model better understand the specific needs of the current zero-shot classification. For example, a certain example input into the large language model can be: "Example 1, the information corresponding to the original material information and the change event information of the credit service can be: the credit service uses a user's face photo for identity verification; the question can be: whether the credit service involves biological identification information, and the output result is yes or no; the answer can be: yes; reason: the face photo belongs to biological identification information."

[0074] In some embodiments, in order to improve the accuracy of zero-shot classification by a large language model, the large language model can be guided to analyze the problem step by step and obtain the final answer. That is, the final question of zero-shot classification can be divided into multiple sub-questions in the order of answers, and the large language model can be guided to answer each sub-question step by step to obtain the final answer. For example, the question of zero-shot classification of a certain credit service can be whether the credit service involves biological identification information, and the corresponding sub-questions can be: first step: identify the data type mentioned in the original data, second step: judge whether the type belongs to biological identification, third step: judge whether the credit service involves biological identification information according to the result of the second step.

[0075] In some embodiments, in order to improve the accuracy of zero-shot classification by a large language model, a large language model supporting structured output can be selected, and the large language model is forced to return JSON (JavaScript Object Notation, a lightweight data exchange format) which has the characteristics of simplicity, readability and easy parsing, so as to ensure that the zero-shot classification result output by the large language model is accurate and convenient for subsequent use.

[0076] Reference Figure 5 A flowchart for generating a credit service portrait provided by an embodiment of the present specification is shown in FIG. 1. As shown in FIG. 1, the credit service portrait generation process includes the following steps: Figure 5 It can be seen that after obtaining the original material information and the change event information of the recommended credit service, the large language model can use the original material information and the change event information to perform semantic extraction, zero-shot classification and knowledge base reasoning, respectively, to obtain second basic feature information, a second classification result and second reasoning feature information. Among them, when performing knowledge base reasoning, the second professional knowledge base corresponding to the recommended credit service needs to be obtained first. Finally, according to the second basic feature information, the second classification result and the second reasoning feature information output by the large language model, the final credit service portrait is obtained.

[0077] In some embodiments of the present specification, the feature mining engine is specifically used for:

[0078] From a plurality of institutions, determine an adjacent institution with the highest coincidence degree of the institution portrait and the target institution portrait;

[0079] Obtain historical data generated by the adjacent institution in the credit service recommendation system;

[0080] According to the historical data generated by the adjacent institution in the credit service recommendation system and the historical data generated by the target institution in the credit service recommendation system, generate prompt words and input a large language model to obtain the intention features of the target institution output by the model.

[0081] In view of the fact that some institutions have a short time of staying in the credit service recommendation system, and thus a small amount of historical data generated. In order to further improve the accuracy of the intention feature of the institution, the institution portrait can be used to determine a neighboring institution with the highest coincidence degree with the target institution portrait from a plurality of institutions, and then the historical data generated by the neighboring institution in the credit service recommendation system is obtained. Finally, the intention feature of the target institution is determined according to the historical data generated by the neighboring institution in the credit service recommendation system and the historical data generated by the target institution in the credit service recommendation system. Since the institution portrait of the neighboring institution is highly similar to that of the target institution, the historical data of the neighboring institution also has a certain reference basis for the target institution, thereby providing more abundant historical data for the target institution and improving the accuracy of the intention feature generation.

[0082] In some embodiments of the present specification, the intelligent recommendation engine is specifically used for:

[0083] obtaining the resource exchange rules of the various candidate credit services;

[0084] determining a target credit service meeting the intention of the target institution from the various candidate credit services according to the resource exchange rules of the various candidate credit services and the intention feature.

[0085] The resource exchange rules of the credit service can include the resources required to exchange a single credit service, and the resources required to exchange a combination of multiple credit services. In some embodiments, the resource exchange rules of the credit service further include differentiated resource exchange preferential treatment provided for different customer levels of different institutions. Since an institution will consider the required exchange resources of different credit services when considering using a credit service, in order to further improve the accuracy of the credit service recommendation, the intelligent recommendation engine in the present embodiment determines a target credit service meeting the intention of the target institution from the various candidate credit services according to the resource exchange rules of the various candidate credit services and the intention feature. It should be noted that the target credit service can be a single credit service or a combination of multiple credit services, which is not limited.

[0086] In some embodiments of the present specification, the intention feature of the target institution includes a current intention feature corresponding to the current demand data and a historical intention feature corresponding to the historical data; and the intelligent recommendation engine is specifically used for:

[0087] determining a target credit service meeting the intention of the target institution from the various candidate credit services according to the resource exchange rules of the various candidate credit services, the current intention feature and the historical intention feature.

[0088] It should be noted that the intention feature of the target institution can simultaneously include a historical intention feature and a current intention feature, wherein the historical intention feature is an intention feature generated by historical data generated by the target institution in the credit service recommendation system, and the current intention feature is an intention feature generated by current demand data input by the target institution. According to the combination of the historical intention feature and the current intention feature, the real intention of the target institution can be more accurately obtained, thereby improving the accuracy of recommending the target credit service for the target institution.

[0089] In some embodiments of the present specification, with reference to Figure 3 , the credit service recommendation system further comprises:

[0090] The negotiation agent 206 is configured to: obtain supplementary explanation information of the target institution, and generate a prompt word according to the supplementary explanation information and input the large language model to update the intention feature to obtain an updated intention feature output by the model; and update the target credit service according to the updated intention feature and the resource exchange rule of each candidate credit service.

[0091] It is considered that the demand of the target institution may not be satisfied by one credit service recommendation, therefore, in the present embodiment, the credit service recommendation system further comprises a negotiation agent, which can negotiate with the institution multiple times about the credit service recommendation to improve the satisfaction of the user. After the intelligent recommendation engine recommends the target credit service to the target institution, if the target institution is not satisfied, the target institution can continue to input supplementary explanation information to the credit service recommendation system, for example, the supplementary explanation information can be "the exchange resource of the currently recommended target credit service is too high, can the credit service with lower exchange resource be recommended". After the target institution inputs the supplementary explanation information, the negotiation agent first obtains the supplementary explanation information, and then generates a prompt word according to the supplementary explanation information and inputs the large language model to update the intention feature to obtain an updated intention feature output by the model. Then, the negotiation agent can update the target credit service according to the updated intention feature and the resource exchange rule of each candidate credit service, that is, update the target credit service to a credit service that meets the updated intention feature. It should be noted that the negotiation agent can update the target credit service for the same target institution multiple times until the target institution no longer inputs new supplementary explanation information.

[0092] In some embodiments of the present specification, with reference to Figure 3 , the credit service recommendation system further comprises:

[0093] The sample test engine 208 is configured to: in response to the negotiation agent updating the target credit service, provide a sample test service corresponding to the updated target credit service to the target institution; in response to the target institution starting the sample test service, obtain test data corresponding to the updated target credit service; run the target credit service according to the test data, and output a running result of the target credit service as reference information for the credit service recommendation system to sign a credit service exchange agreement with the target institution.

[0094] To better help the target institution to understand the performance of the target credit service in advance, thereby providing a basis for the target credit institution to sign a credit service exchange agreement, in the embodiments of the present specification, the credit service recommendation system further comprises a sample test engine. After the negotiation agent updates the target credit service, the sample test engine provides a sample test service corresponding to the updated target credit service to the target institution. If the target institution starts the sample test service, the sample test engine first obtains test data corresponding to the updated target credit service. The test data can generally be provided by the target institution to ensure the fairness of the test result. In some embodiments, the test data can also be provided by the sample test engine, which is not limited. After obtaining the test data, the sample test engine can run the target credit service according to the test data, and output a running result of the target credit service as reference information for the credit service recommendation system to sign a credit service exchange agreement with the target institution. In some embodiments, while outputting the running result of the target credit service, the credit service recommendation system can also receive feedback opinions of the target institution on the test result, which can be used to further optimize the credit service. If the feedback opinions of the target institution indicate that the target institution is satisfied with the test result, the credit service recommendation system can further provide a window for the target institution to sign a credit service exchange agreement, and the target institution can complete the signing of the credit service exchange agreement through the window.

[0095] Reference Figure 6 A schematic diagram of a credit product recommendation process provided by the embodiments of the present specification, in some embodiments, Figure 6 The credit product in the credit product recommendation process is a credit service, such as Figure 6As shown, when the institution accesses the credit service recommendation system, the institution portrait can be completed according to the original data provided by the institution when accessing, and the institution portrait library is formed by the institution quality of a plurality of institutions that have accessed. At the same time, when the credit service to be recommended is registered in the credit service recommendation system, the product portrait (i.e., the credit service portrait) can be completed according to the original materials provided by the product when registering, and the product portrait library is formed by a plurality of credit services to be recommended. When the characteristics of the institution are mined, the intention characteristics and the correlation of the institution are generated by the feature mining engine through the correlation data generated by the institution in the credit service recommendation system, the institution portrait library and the product portrait library, and then the intelligent recommendation engine recommends the target credit service for the institution according to the intention characteristics and the correlation, and displays the recommendation result to the institution. After the institution receives the target credit service, the exchange resource negotiation can be performed through the negotiation intelligent agent, and the recommendation result is adjusted according to the result of the exchange resource agreement, so that the adjusted recommendation result better meets the needs of the user. After the exchange resource negotiation result, the sample test service can be provided for the institution, and the credit service exchange agreement is signed in the case that the institution is satisfied with the sample test result.

[0096] The credit service recommendation system of the embodiment of the present specification, the feature mining engine generates prompt words according to the target institution portrait of the target institution and the credit service portrait of each credit service to be recommended and inputs a large language model to obtain the target correlation between the target institution portrait and each credit service portrait output by the model; and generates prompt words according to the correlation data corresponding to the target institution and inputs a large language model to obtain the intention characteristics of the target institution output by the model; the feature mining engine can accurately mine the correlation between the institution portrait and the credit service portrait, so as to obtain the preference of different institutions for credit services. At the same time, the feature mining engine can also determine the real intention of the institution according to the correlation data corresponding to the institution, which provides a basis for the intelligent recommendation engine to accurately recommend the target credit service. The intelligent recommendation engine determines the target credit service that meets the final needs of the institution from each credit service to be recommended according to the correlation and the intention characteristics corresponding to the institution. Since the preference of the institution for each credit service and the real needs of the institution are fully considered, the intelligent recommendation engine improves the accuracy of recommending the target credit service for the institution. In addition, the credit service recommendation system provided by the embodiment of the present specification proposes a double-portrait system of institution portrait + credit service portrait, and constructs the correlation between the institution and the credit service according to the double-portrait system, which becomes the basis for accurate matching of credit services and supports high-dimensional heterogeneous data fusion. Moreover, according to the automatic mining mechanism of the characteristics of the large language model, end-to-end conversion from unstructured demand text to computable features is realized.

[0097] Corresponding to the above credit service recommendation system, the present specification also proposes a credit service recommendation method in the embodiment, which is applied to the intelligent recommendation engine, and the credit service recommendation method comprises the following steps of Figure 7The method comprises the following steps:

[0098] S702, in response to a credit service recommendation demand issued by a target institution to a credit service recommendation system, determining each candidate credit service matched with the target institution from each credit service to be recommended according to a target correlation corresponding to the target institution; the target correlation is a correlation between a target institution portrait and each credit service portrait obtained by a feature mining engine by generating a prompt word and inputting a large language model according to the target institution portrait of the target institution and the credit service portrait of each credit service to be recommended;

[0099] S704, determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the intention feature of the target institution, so as to recommend the target credit service to the target institution; the intention feature is obtained by the feature mining engine by generating a prompt word and inputting a large language model according to the correlation data corresponding to the target institution; the correlation data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution.

[0100] In some embodiments of the present specification, determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the intention feature of the target institution comprises:

[0101] Obtaining resource exchange rules of the candidate credit services;

[0102] Determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services and the intention feature.

[0103] In some embodiments of the present specification, the intention feature of the target institution includes a current intention feature corresponding to the current demand data and a historical intention feature corresponding to the historical data; determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services and the intention feature comprises:

[0104] Determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services, the current intention feature and the historical intention feature.

[0105] In some embodiments of the present specification, the process in which the feature mining engine determines the target institutional portrait of the target institution includes: obtaining an institution corresponding to each institutional portrait in a plurality of pre-generated institutional portraits; in response to determining that the institution corresponding to the each institutional portrait does not include the target institution, determining a replaceable institution matching the target institution from the institution corresponding to the each institutional portrait according to attribute information of the target institution, and determining an institutional portrait corresponding to the replaceable institution as the target institutional portrait of the target institution.

[0106] In some embodiments of the present specification, the method further includes:

[0107] For each institution for which an institutional portrait is generated, obtaining raw data information of the institution and new event information related to the institution;

[0108] Performing semantic analysis on the raw data information and the new event information by a large language model, and obtaining first basic feature information of the institution according to a first result of the semantic analysis;

[0109] Performing zero-shot classification on the institution according to the raw data information and the new event information, and obtaining a first classification result of the institution;

[0110] Obtaining a first professional knowledge base corresponding to the institution, and obtaining first reasoning feature information of the institution according to the first professional knowledge base, the raw data information and the new event information;

[0111] Determining an institutional portrait of the institution according to the first basic feature information, the first classification result and the first reasoning feature information.

[0112] In some embodiments of the present specification, the method further includes:

[0113] For each of the various recommended credit investigation services, obtaining raw material information of the recommended credit investigation service and change event information related to the recommended credit investigation service;

[0114] Performing semantic analysis on the raw material information and the change event information by a large language model, and obtaining second basic feature information of the recommended credit investigation service according to a second result of the semantic analysis;

[0115] Performing zero-shot classification on the recommended credit investigation service according to the raw material information and the change event information, and obtaining a second classification result of the recommended credit investigation service;

[0116] obtain a second professional knowledge base corresponding to the to-be-recommended credit service, and obtain second reasoning feature information of the to-be-recommended credit service according to the second professional knowledge base, the original material information, and the change event information;

[0117] obtain a credit service portrait of the to-be-recommended credit service according to the second basic feature information, the second classification result, and the second reasoning feature information.

[0118] In some embodiments of the present specification, the process in which the feature mining engine determines the intention feature of the target institution includes: determining, from a plurality of institutions, a neighboring institution with the highest coincidence degree of the institution portrait and the target institution portrait; obtaining historical data generated by the neighboring institution in the credit service recommendation system; generating a prompt word according to the historical data generated by the neighboring institution in the credit service recommendation system and the historical data generated by the target institution in the credit service recommendation system, and inputting the large language model to obtain the intention feature of the target institution output by the model.

[0119] In some embodiments of the present specification, the method further includes:

[0120] obtaining supplementary explanation information of the target institution, and updating the intention feature according to the supplementary explanation information by generating a prompt word and inputting the large language model to obtain the updated intention feature output by the model;

[0121] updating the target credit service according to the updated intention feature and the resource exchange rule of each alternative credit service.

[0122] In some embodiments of the present specification, the method further includes:

[0123] In response to updating the target credit service by the negotiation intelligent agent, providing a sample test service corresponding to the updated target credit service to the target institution;

[0124] In response to the target institution starting the sample test service, obtaining test data corresponding to the updated target credit service;

[0125] running the target credit service according to the test data, and outputting a running result of the target credit service as reference information for the credit service recommendation system and the target institution to sign a credit service exchange agreement.

[0126] The credit service recommendation method of the above embodiments is used to implement the functions of the corresponding credit service recommendation system in any of the preceding embodiments, and has the beneficial effects of the corresponding credit service recommendation system embodiments, which will not be repeated here.

[0127] Figure 8 This is a schematic structural diagram of a device provided in an exemplary embodiment. For example... Figure 8 As shown, device 800 mainly consists of a communication interface 802, a mechanism interface 804, a processor 806, and a data storage 808. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 810. The communication interface 802 enables device 800 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 802 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 802 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 802 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 802 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0128] Mechanism interface 804 includes receiving mechanism input and providing output to the mechanism. Therefore, mechanism interface 804 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. Mechanism interface 804 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, mechanism interface 804 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external mechanism input / output devices. Additionally or alternatively, device 800 may support remote access from other devices via communication interface 802 or another physical interface (not shown). Mechanism interface 804 may be configured to receive mechanism input, the position and movement of which may be indicated by indicators or cursors described herein. Mechanism interface 804 may also be configured as a display device for rendering or displaying text fragments.

[0129] The processor 806 may contain one or more general-purpose processors and / or special-purpose processors.

[0130] Data storage 808 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 806. Data storage 808 may include removable and non-removable components.

[0131] The processor 806 is capable of executing program instructions stored in the data storage 808 (e.g., compiled or interpreted program logic and / or machine code) to implement various functionality described herein. The data storage 808 can include a non-transitory computer-readable medium having stored thereon program instructions that, when executed by the device 800, enable the device 800 to carry out any of the systems, methods or functions disclosed in the specification and / or drawings. The execution of the program instructions 818 by the processor 806 can cause the processor 806 to utilize the data 812.

[0132] For example, the program instructions 818 can include an operating system 822 (e.g., an operating system kernel, device drivers, and / or other modules) installed on the device 800 as well as one or more application programs 820 (e.g., a browser, a social application, or a gaming application). Similarly, the data 812 can include operating system data 816 and application data 814. The operating system data 816 is primarily accessible to the operating system 822, while the application data 814 is primarily accessible to the one or more application programs 820. The application data 814 can be located in a file system that is visible or hidden from the mechanisms of the device 800.

[0133] The application programs 820 can communicate with the operating system 822 through one or more application programming interfaces (APIs). These APIs facilitate the application programs 820 reading and / or writing to the application data 814, transmitting or receiving information via the communication interface 802, receiving or displaying information on the mechanism interface 804, and the like.

[0134] In some terminology, the application programs 820 can be referred to simply as “apps.” Furthermore, the application programs 820 can be downloaded to the device 800 through one or more online application stores or application markets. However, the application programs can also be installed on the device 800 through other means, such as through a web browser or a physical interface (e.g., a USB port) on the device 800.

[0135] Reference is made to Figure 2 , the credit investigation service recommendation system can be applied to the device as shown in Figure 8 to implement the technical solutions of the specification.

[0136] For ease of description, the above system is described in various modules or units in terms of functions. Of course, the functions of each module or unit can be implemented in the same or more software and / or hardware when implementing one or more of the present specification, or the modules implementing the same function can be implemented by a combination of multiple sub-modules or sub-units. The above described device embodiments are only illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division method, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0137] According to the same concept as the above system, the present specification also provides an electronic device, comprising: a processor; a memory for storing processor executable instructions; wherein the processor implements the steps of the credit service recommendation method according to any one of the above embodiments by running the executable instructions.

[0138] According to the same concept as the above system, the present specification also provides a computer readable storage medium, having computer instructions stored thereon, which are executed by a processor to implement the steps of the credit service recommendation method according to any one of the above embodiments.

[0139] According to the same concept as the above system, the present specification also provides a computer program product, comprising computer program / instructions, which are executed by a processor to implement the steps of the credit service recommendation method according to any one of the above embodiments.

[0140] Those skilled in the art can understand that:

[0141] In the present specification, the term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, system, product or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or further includes elements inherent to such a process, method, system, product or device. Without more limitations, it does not exclude the presence of other same or equivalent elements in the process, system, product or device including the elements.

[0142] In the present specification, "one", "a" and "the" do not refer to a single number, but also include plural.

[0143] In the present specification, the first, second, etc. ordinal numbers do not necessarily indicate the order, and many times are used for the purpose of distinguishing the objects. For example, the first server and the second server usually refer to two servers. In order to distinguish the two servers, they are expressed as the first server and the second server. Of course, at times, the two servers can also be the same server.

[0144] In this specification, unless specifically stated otherwise, "receiving, sending" of data is not necessarily direct receiving and sending, and can be indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving data sent by B, or can be understood as A indirectly receiving data sent by B through C or other subject. Similarly, B sending data to A can be understood as B directly sending data to A, or can be understood as B indirectly sending data to A through C or other subject. Here, C can be one subject, or two or more subjects.

[0145] In this specification, unless specifically stated otherwise, the association relationship generated between structures can be a direct association relationship or an indirect association relationship. For example, when "A is connected with B" is described, unless it is specifically stated that A is directly connected with B, it should be understood that A can be directly connected with B, or A can be indirectly connected with B; for another example, when "A is on B" is described, unless it is specifically stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B, or A can be indirectly above B (there are other elements between AB, and A is above B). By analogy.

[0146] This specification uses specific words to describe the embodiments of this specification. As "one embodiment", "an embodiment", and / or "some embodiments" means a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that the "an embodiment" or "one embodiment" or "one alternative embodiment" mentioned in this specification at different positions two or more times does not necessarily refer to the same embodiment. In addition, the skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples without contradiction.

[0147] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it can be understood that the order of steps listed in the embodiments or flowcharts is only one of the many execution orders, and does not represent the only execution order. Therefore, when the claims involve method steps, the changes and adjustments of the order of the steps or the parallelism between the steps are also within the scope of protection of the claims.

Claims

1. A credit service recommendation system, comprising: a feature mining engine configured to generate a prompt word and input a large language model according to a target institution portrait of a target institution and a credit service portrait of each credit service to be recommended, to obtain a target association relationship between the target institution portrait and each credit service portrait output by the model, and to generate a prompt word and input a large language model according to association data corresponding to the target institution, to obtain an intention feature of the target institution output by the model; the association data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution; an intelligent recommendation engine configured to determine each candidate credit service matching the target institution from the each credit service to be recommended according to the target association relationship, and to determine a target credit service meeting the intention of the target institution from the each candidate credit service according to the intention feature, to recommend the target credit service to the target institution; wherein the system further comprises an institution portrait engine configured to: for each institution for which an institution portrait is generated, obtain original data information of the institution and new event information related to the institution; perform semantic analysis on the original data information and the new event information by a large language model, and obtain first basic feature information of the institution according to a first result of the semantic analysis; perform zero-shot classification on the institution according to the original data information and the new event information, to obtain a first classification result of the institution; obtain a first professional knowledge base corresponding to the institution, and obtain first reasoning feature information of the institution according to the first professional knowledge base, the original data information and the new event information; determine an institution portrait of the institution according to the first basic feature information, the first classification result and the first reasoning feature information.

2. The system of claim 1, wherein the feature mining engine is further configured to: obtain institutions corresponding to each institution portrait in a plurality of pre-generated institution portraits; in response to determining that the institutions corresponding to the each institution portrait do not include the target institution, determine a replaceable institution matching the target institution from the institutions corresponding to the each institution portrait according to attribute information of the target institution; determine an institution portrait corresponding to the replaceable institution as a target institution portrait of the target institution.

3. The system of claim 1, wherein the system further comprises a credit service portrait engine configured to: for each credit service to be recommended in the each credit service to be recommended, obtain original material information of the credit service to be recommended and change event information related to the credit service to be recommended; perform semantic analysis on the original material information and the change event information by a large language model, and obtain second basic feature information of the credit service to be recommended according to a second result of the semantic analysis; perform zero-shot classification on the credit service to be recommended according to the original material information and the change event information, to obtain a second classification result of the credit service to be recommended; obtaining a second professional knowledge base corresponding to the to-be-recommended credit investigation service, and obtaining second reasoning feature information of the to-be-recommended credit investigation service according to the second professional knowledge base, the original material information, and the change event information; obtaining a credit investigation service portrait of the to-be-recommended credit investigation service according to the second basic feature information, the second classification result, and the second reasoning feature information.

4. The system of claim 1, wherein the feature mining engine is specifically configured to: determine, from a plurality of institutions, a neighboring institution having the highest coincidence degree with the target institution portrait; obtain historical data generated by the neighboring institution in the credit investigation service recommendation system; generate prompt words based on the historical data generated by the neighboring institution in the credit investigation service recommendation system and the historical data generated by the target institution in the credit investigation service recommendation system, and input the prompt words into a large language model to obtain an intention feature of the target institution output by the model.

5. The system of claim 1, wherein the intelligent recommendation engine is specifically configured to: obtain resource exchange rules of the respective candidate credit investigation services; determine a target credit investigation service meeting the intention of the target institution from the respective candidate credit investigation services according to the resource exchange rules of the respective candidate credit investigation services and the intention feature.

6. The system of claim 5, wherein the intention feature of the target institution includes a current intention feature corresponding to the current demand data and a historical intention feature corresponding to the historical data; and the intelligent recommendation engine is specifically configured to: determine a target credit investigation service meeting the intention of the target institution from the respective candidate credit investigation services according to the resource exchange rules of the respective candidate credit investigation services, the current intention feature, and the historical intention feature.

7. The system of claim 5 or 6, further comprising: a negotiation intelligent agent configured to: obtain supplementary description information of the target institution, generate prompt words based on the supplementary description information, input the prompt words into a large language model to update the intention feature, and obtain an updated intention feature output by the model; and update the target credit investigation service according to the updated intention feature and the resource exchange rules of the respective candidate credit investigation services.

8. The system of claim 7, further comprising: a sample test engine configured to: in response to updating the target credit investigation service by the negotiation intelligent agent, provide a sample test service corresponding to the updated target credit investigation service to the target institution; and in response to the target institution starting the sample test service, obtain test data corresponding to the updated target credit investigation service; run the target credit investigation service according to the test data, and output a running result of the target credit investigation service as reference information for signing a credit investigation service exchange agreement between the credit investigation service recommendation system and the target institution.

9. A credit investigation service recommendation method, the method being applied to an intelligent recommendation engine, and the method comprising: ​ In response to a credit service recommendation demand issued by a target institution to a credit service recommendation system, each candidate credit service matched with the target institution is determined from each credit service to be recommended according to a target association relationship corresponding to the target institution; The target association relationship is an association relationship between a target institution portrait and each credit service portrait obtained by a feature mining engine generating a prompt word and inputting a large language model according to the target institution portrait and the credit service portrait of each credit service to be recommended; According to the intention feature of the target institution, a target credit service conforming to the intention of the target institution is determined from the candidate credit services, so as to recommend the target credit service to the target institution; the intention feature is obtained by the feature mining engine generating a prompt word and inputting a large language model according to the association data corresponding to the target institution; the association data includes historical data generated by the target institution in the credit service recommendation system and / or current demand data input by the target institution; The method further comprises: For each institution generating an institution portrait, obtain the original data information of the institution and the new event information related to the institution; Perform semantic analysis on the original data information and the new event information by a large language model, and obtain first basic feature information of the institution according to a first result of the semantic analysis; Classify the institution according to the original data information and the new event information to obtain a first classification result of the institution; Obtain a first professional knowledge base corresponding to the institution, and obtain first reasoning feature information of the institution according to the first professional knowledge base, the original data information and the new event information; Determine the institution portrait of the institution according to the first basic feature information, the first classification result and the first reasoning feature information.

10. The method of claim 9, wherein determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the intention feature of the target institution comprises: Obtaining resource exchange rules of the candidate credit services; Determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services and the intention feature.

11. The method of claim 10, wherein the intention feature of the target institution includes a current intention feature corresponding to the current demand data and a historical intention feature corresponding to the historical data; and determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services and the intention feature comprises: Determining a target credit service conforming to the intention of the target institution from the candidate credit services according to the resource exchange rules of the candidate credit services, the current intention feature and the historical intention feature.

12. An electronic device comprising: a processor; a memory for storing processor-executable instructions; wherein the processor implements the method of any one of claims 9-11 by executing the executable instructions.

13. A computer readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the steps of the method of any one of claims 9-11.

14. A computer program product, comprising: Computer program / instructions that, when executed by a processor, implement the method of any one of claims 9-11.

Images