Landing query request processing method and device, equipment, medium and program product

Abstract

The embodiment of the invention provides a loan query request processing method and device, equipment, a medium and a program product. The method comprises the steps of obtaining a loan query request sent by a loan terminal of a user, processing the loan query request to obtain a query problem vector, determining a plurality of matched knowledge block vectors in a vector database according to the query problem vector to obtain a retrieval result, and sending the retrieval result to the loan terminal of the user. Then, according to user information and question types corresponding to the query question vectors, knowledge block vectors in the retrieval results are updated to obtain updated retrieval results, reordering is achieved by re-determining the similarity of all the knowledge block vectors in the updated retrieval results, and target retrieval results are obtained; and generating request data according to the target retrieval result and feeding back the request data to the user loan terminal. The method is used for achieving the effects of automatically obtaining the request data corresponding to the query problem, improving the accuracy of the obtained request data, improving the loaning efficiency and enhancing the user experience.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to a method, device, equipment, medium and program product for processing loan inquiry requests. Background Art

[0002] At present, the loan business is highly complex, and the bank loan operation procedures and precautions for different products and / or different links are different.

[0003] In the prior art, if an account manager encounters an unclear step in the process of handling a loan application, the account manager needs to leave the current process operation page and obtain information from internal regulations or operation manuals through the query entrance of the regulations or operation manuals to determine the subsequent operation steps (or when the system reports an error, obtain the system repair method by asking the relevant developers).

[0004] However, the existing technology frequently switches interfaces to query the operation process, which reduces the loan efficiency, and asking developers increases the communication time cost during business processing, prolongs the loan time, and further reduces the loan efficiency. Summary of the invention

[0005] The embodiments of the present application provide a method, device, equipment, medium and program product for processing loan query requests, so as to reduce loan costs, improve loan efficiency and enhance user experience.

[0006] In a first aspect, an embodiment of the present application provides a method for processing a loan inquiry request, which is applied to a loan inquiry system, comprising:

[0007] Obtaining a loan query request sent by a user's loan terminal, processing the loan query request, and obtaining a query question vector, wherein the loan query request includes a query question and user information;

[0008] Determine the search result corresponding to the query question based on the vector database, the search result includes multiple knowledge block vectors; obtain updated search results according to the user information and the question category corresponding to the query question;

[0009] Determine the similarity between each knowledge block vector in the updated retrieval result and the query question, and sort the multiple knowledge block vectors based on the obtained similarities to obtain the target retrieval result; generate prompt data corresponding to the loan query request based on the target retrieval result, determine the corresponding request data based on the prompt data and feed it back to the user loan terminal.

[0010] Optionally, before determining the search result, a material data set updated in real time is obtained, where the material data set includes a plurality of material data and material tag information corresponding to each material data;

[0011] By preprocessing the material data set, a plurality of text data in a target format are obtained, and the text type of each text data is determined;

[0012] For any text data type, the corresponding segmentation strategy is determined based on the text type, and the text data is segmented based on the segmentation strategy to obtain multiple text knowledge blocks corresponding to the text data. The multiple text knowledge blocks are vectorized to obtain multiple knowledge block vectors corresponding to the text data and stored in the vector database.

[0013] Optionally, based on the user information and the question category corresponding to the query question, an updated search result is obtained, specifically including:

[0014] Determine label information of each knowledge block vector in the retrieval result based on the material label information;

[0015] Determine the similarity between each tag information in the search result and the user information to obtain a first knowledge block vector set, and determine the similarity between each tag information and the question category to obtain a second knowledge block vector set;

[0016] Based on the weight mapping table, determine the first weight data corresponding to the user information and the second weight data corresponding to the question category; perform weighted processing on the first knowledge block vector set according to the first weight data, and perform weighted processing on the second knowledge block vector set according to the second weight data;

[0017] The updated retrieval result is determined based on the weighted knowledge block vector.

[0018] Optionally, obtain real-time business information of the user's loan terminal;

[0019] A target knowledge block vector is determined based on real-time business information, and whether the real-time business information is valid is determined based on the target knowledge block vector; and when it is determined to be invalid, an alarm indication is generated based on the invalid business information and fed back to the user's loan terminal.

[0020] Optionally, user query data is generated and sent to the audit terminal, so that the audit terminal determines whether the user query data is data to be revised, and generates update data when it is determined to be data to be revised;

[0021] Receive the update data sent by the audit terminal, and update the vector database according to the update data.

[0022] Optionally, obtaining a historical loan query record from the user, the historical loan query record including a plurality of historical loan query requests;

[0023] When it is determined that there is an update of the knowledge block vector corresponding to any historical loan query request, historical update request data corresponding to the historical loan query request is generated based on the updated knowledge block vector and fed back to the user loan terminal.

[0024] In a second aspect, an embodiment of the present application provides a processing device for a loan inquiry request, which is applied to a loan inquiry system, including:

[0025] An acquisition module is used to acquire a loan query request sent by a user's loan terminal, process the loan query request, and obtain a query question vector. The loan query request includes a query question and user information.

[0026] A processing module, used to determine the search result corresponding to the query question based on the vector database, the search result including multiple knowledge block vectors; and obtain updated search results according to the user information and the question category corresponding to the query question;

[0027] The processing module is also used to determine the similarity between each knowledge block vector in the updated retrieval results and the query question, and sort multiple knowledge block vectors based on the obtained similarities to obtain the target retrieval results; generate prompt data corresponding to the loan query request based on the target retrieval results, determine the corresponding request data based on the prompt data and feed it back to the user loan terminal.

[0028] Optionally, the acquisition module is further used to acquire a material data set updated in real time before determining the search result, the material data set including a plurality of material data and material tag information corresponding to each material data;

[0029] By preprocessing the material data set, a plurality of text data in a target format are obtained, and the text type of each text data is determined;

[0030] For any text data type, the corresponding segmentation strategy is determined based on the text type, and the text data is segmented based on the segmentation strategy to obtain multiple text knowledge blocks corresponding to the text data. The multiple text knowledge blocks are vectorized to obtain multiple knowledge block vectors corresponding to the text data and stored in the vector database.

[0031] Optionally, the processing module is further used to obtain updated search results according to the user information and the question category corresponding to the query question, specifically including:

[0032] Determine label information of each knowledge block vector in the retrieval result based on the material label information;

[0033] Determine the similarity between each tag information in the search result and the user information to obtain a first knowledge block vector set, and determine the similarity between each tag information and the question category to obtain a second knowledge block vector set;

[0034] Based on the weight mapping table, determine the first weight data corresponding to the user information and the second weight data corresponding to the question category; perform weighted processing on the first knowledge block vector set according to the first weight data, and perform weighted processing on the second knowledge block vector set according to the second weight data;

[0035] The updated retrieval result is determined based on the weighted knowledge block vector.

[0036] Optionally, the acquisition module is also used to obtain real-time business information of the user's loan terminal;

[0037] A target knowledge block vector is determined based on real-time business information, and whether the real-time business information is valid is determined based on the target knowledge block vector; and when it is determined to be invalid, an alarm indication is generated based on the invalid business information and fed back to the user's loan terminal.

[0038] Optionally, the processing module is further used to generate user query data and send it to the audit terminal, so that the audit terminal determines whether the user query data is data to be revised, and generates update data when it is determined to be data to be revised;

[0039] Receive the update data sent by the audit terminal, and update the vector database according to the update data.

[0040] Optionally, the acquisition module is further used to acquire a historical loan query record of the user, where the historical loan query record includes multiple historical loan query requests;

[0041] When it is determined that there is an update of the knowledge block vector corresponding to any historical loan query request, historical update request data corresponding to the historical loan query request is generated based on the updated knowledge block vector and fed back to the user loan terminal.

[0042] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0043] Memory stores computer-executable instructions;

[0044] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0045] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0046] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0047] The processing method, device, equipment, medium and program product of the loan query request provided in the embodiment of the present application obtains the loan query request sent by the user's loan terminal, processes the loan query request, obtains the query question vector, and determines multiple matching knowledge block vectors in the vector database according to the query question vector to obtain the search result. Then, according to the user information and question category corresponding to the query question vector, the knowledge block vector in the search result is updated to obtain the updated search result. The similarity of each knowledge block vector in the updated search result is re-determined to achieve re-sorting to obtain the target search result, thereby generating request data according to the target search result and feeding it back to the user's loan terminal. The present application realizes the automatic acquisition of request data corresponding to the query question, and improves the accuracy of the obtained request data, improves the loan efficiency, and enhances the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0049] Figure 1 A schematic diagram of the prior art provided for this application;

[0050] Figure 2 A flowchart of a method for processing a loan inquiry request provided for this application;

[0051] Figure 3 A schematic diagram of the structure of the loan inquiry system provided for this application;

[0052] Figure 4 A schematic diagram of the structure of a processing device for a loan inquiry request provided by the present application;

[0053] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.

[0054] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0055] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0056] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0057] In addition, this application involves conducting big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and using artificial intelligence technology to make automated decisions, and making technical solutions that have a significant impact on personal rights and interests based on the results of automated decisions. The application provides users with corresponding operation entrances for them to choose to agree or reject the results of automated decisions; if the user chooses to reject, the expert decision-making process will be entered.

[0058] At present, bank loan business is highly complex. Different products (such as discount, letter of guarantee, bank acceptance, fixed loan, floating loan, etc.) and different links (such as credit limit effectiveness, application maintenance and signing, scanning and filing, review and approval, etc.) have different operation procedures and precautions. When handling loan business, account managers are usually required to handle business based on various internal bank regulations, management methods and system operation manuals. Therefore, loan business has strict requirements on account managers' professional knowledge, proficiency and operational capabilities. In addition, due to the high mobility of account managers in bank branches or online loan companies in different regions, the process proficiency and operational proficiency of loan system in the process of loan business are reduced.

[0059] In order to solve the above problems, the prior art provides auxiliary assistance to the account manager through ES retrieval. Figure 1 The schematic diagram of the prior art provided for this application is as follows: Figure 1As shown, the loan side includes loan terminals 1, ..., loan terminals n. When the account manager handles the loan business on any loan terminal (such as loan terminal 1), he needs to interact with the loan terminal to transmit the relevant business information to the loan system for approval and the next step. Due to the complexity of the business process, when handling urgent or difficult loan business, even experienced account managers may encounter uncertain or unclear steps. At this time, the account manager needs to send query information to the OA library through the regulations query interface of loan terminal 1 to obtain relevant information from the OA library, so as to further carry out the subsequent business processing process.

[0060] However, when the prior art encounters the need to query internal regulations and operating manual materials during the operation process, it is necessary to switch from the business processing interface of the loan terminal to the regulations query interface, and browse the query results when the query results are obtained, and then reopen the business processing page to perform subsequent business processing based on the remembered query results, thereby increasing time and labor costs and reducing business processing efficiency.

[0061] In addition, with the continuous iteration and update of information, the revision and replacement of new products, new processes and new regulations within the bank are relatively frequent, and the data knowledge base content of the OA library is mostly general knowledge, which fails to focus highly on a certain vertical field and is difficult to be updated in a timely manner, resulting in poor accuracy of the query results fed back by the OA library. Therefore, it is difficult for the existing technology to provide the account manager with relevant regulatory data, the system's standard operating procedures, and the solution to the system error, which easily leads to the account manager spending a lot of time to query relevant information or ask management personnel or developers (such as through emails, telephone communications, etc.), which greatly increases the communication cost, prolongs the loan time, and affects the overall efficiency. In addition, during busy business periods, frequent interruptions to the business processing process are likely to increase the time and labor costs of account managers and loan review staff, thereby affecting customer service quality and customer satisfaction.

[0062] The processing method of the loan query request provided by the present application receives a material data set in real time, and performs segmentation processing on the material data set with different segmentation strategies, obtains multiple knowledge block vectors and stores them in a vector database, after obtaining the loan query request, determines the search result related to the query question vector corresponding to the loan query request from the vector database, and then determines the knowledge block vector corresponding to the user information of the query question vector among the multiple knowledge block vectors in the search result, and performs weighted processing on the knowledge block vector corresponding to the filtered user information according to the first weight data mapped by the user information, and at the same time, determines the knowledge block vector corresponding to the question category of the query question vector, and performs weighted processing on the knowledge block vector corresponding to the filtered question category according to the second weight data mapped by the question category, sorts the similarity of the weighted knowledge block vectors, obtains the target search result, and generates the prompt data corresponding to the loan query request based on the target search result, and feeds back the request data corresponding to the prompt data to the user loan terminal. The present application improves the accuracy of the request data fed back by the user loan terminal, improves the feedback efficiency, improves the operational convenience of the business handling process, reduces the time cost of the loan process, improves the loan efficiency, and enhances the user experience.

[0063] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0064] Figure 2 A flowchart of a method for processing a loan inquiry request provided for this application is shown in FIG. Figure 2 As shown, the method includes:

[0065] S201. Obtain a loan query request, process the loan query request, and obtain a query question vector.

[0066] More specifically, a loan query request sent by a user's loan terminal is obtained, and the loan query request is processed to obtain a query question vector, wherein the loan query request includes a query question and user information. The user refers to a staff member who operates the user's loan terminal, for example, a loan account manager.

[0067] Exemplarily, the user information includes, but is not limited to, the account manager's job role, authority information, current business handling area, and other information.

[0068] Optionally, before determining the retrieval results, a material data set that is updated in real time is obtained, the material data set including multiple material data and material tag information corresponding to each material data; multiple text data in a target format are obtained by preprocessing the material data set, and the text type of each text data is determined; for the text type of any text data, a corresponding segmentation strategy is determined based on the text type, and the text data is segmented based on the segmentation strategy to obtain multiple text knowledge blocks corresponding to the text data, and the multiple text knowledge blocks are vectorized to obtain multiple knowledge block vectors corresponding to the text data and store them in a vector database.

[0069] In a possible embodiment, the material data includes but is not limited to loan regulations, loan business management methods, and loan system operation procedures. The material tag information of each material data includes but is not limited to the publication time, publication organization, applicable job role, and authorized information of the material data.

[0070] Exemplarily, the text type is determined based on metadata of the text data, where the metadata includes but is not limited to the file name and chapter title of the text data.

[0071] Exemplarily, segmentation methods include but are not limited to sentence segmentation, fixed segmentation, and intention segmentation.

[0072] In a possible embodiment, a material data set that is updated and stored in real time is obtained based on the knowledge base of the loan query system. The material data set in the knowledge base includes, but is not limited to, material data regularly updated in the OA library and material data uploaded by users through user loan terminals. Before uploading the data, it is necessary to review the material data for compliance through the review terminal, and add identification information to each material data after the review is passed, so that the material data set with the added identification information is sent to the knowledge base of the loan query system for storage.

[0073] This embodiment obtains a real-time updated material data set from a knowledge base, and after obtaining the material data set, adopts different segmentation methods to perform data segmentation on different types of text data therein, thereby achieving timely and accurate segmentation of the real-time updated text data, and stores it in the vector database after vectorization, thereby improving the accuracy of data segmentation, improving the reliability of the vector database, and further improving the accuracy and reliability of the request data finally generated.

[0074] S202: Determine the search result corresponding to the query question based on the vector database; and obtain updated search results according to the user information and the question category of the query question.

[0075] More specifically, the search results corresponding to the query question are determined based on the vector database, and the search results include multiple knowledge block vectors; and updated search results are obtained according to the user information and the question category corresponding to the query question. The question categories include but are not limited to process operation category, regulation text category, system repair category, etc.

[0076] In a possible embodiment, when a branch account manager queries "application conditions for personal emergency loans" through a user loan terminal, the query question is vectorized to obtain a query question vector, and a search scope that meets the branch account manager's authority is determined in a vector database based on the query question vector and the branch account manager's user information, and a knowledge block vector related to the application conditions for personal emergency loans is searched within the search scope, and the multiple queried knowledge block vectors are determined as the search results corresponding to the query question vector. This embodiment achieves the goal of not feeding back text data that exceeds the branch account manager's authority to the user loan terminal, thereby reducing the possibility of internal bank information being viewed or abused by unauthorized personnel.

[0077] In a possible embodiment, after the search results are obtained, the material upload time corresponding to each knowledge block vector in the search results is determined, and multiple knowledge block vectors are sorted according to the material upload time, so that the knowledge block vector corresponding to the material data of the most recent date is arranged first and presented to the user. By sorting the knowledge block vectors, the efficiency of users obtaining useful data is improved, the efficiency of users handling business is further improved, and the user experience and customer experience are enhanced.

[0078] Illustratively, after the search results are found, the embodiments of the present application may also sort the knowledge block vectors in the search results according to job roles, current business processing areas, and other conditions, and the embodiments of the present application do not impose any restrictions on this.

[0079] In a possible embodiment, the question category of the query question vector is determined to be a rule text category, and the user information of the query question is determined to be the corresponding user's job role, authority information, and current business handling area, and the search results are updated according to the rule text category, job role, authority information, and business handling area to obtain updated search results. For example, the updated search results include 10 knowledge block vectors, and each knowledge block vector corresponds to a similarity value with the query question vector.

[0080] Optionally, the question category and user information of the query question in this embodiment can be determined by obtaining the identifier added when the query question is initiated. The method of determining based on the added identifier is a prior art and will not be elaborated in this embodiment.

[0081] Optionally, based on the user information and the question category corresponding to the query question, an updated retrieval result is obtained, specifically including: determining the label information of each knowledge block vector in the retrieval result based on the material label information; determining the similarity between each label information in the retrieval result and the user information to obtain a first knowledge block vector set, and determining the similarity between each label information and the question category to obtain a second knowledge block vector set; based on a weight mapping table, determining the first weight data corresponding to the user information and the second weight data corresponding to the question category; weighting the first knowledge block vector set according to the first weight data, and weighting the second knowledge block vector set according to the second weight data; determining the updated retrieval result based on the weighted knowledge block vectors.

[0082] In a possible embodiment, the label information corresponding to each knowledge block vector in the retrieval result is determined, and the knowledge block vectors (e.g., 2) belonging to the regulation text class are determined based on multiple label information, and the first weight data corresponding to the problem category of the regulation text class is determined, and the first weight data is weighted for the similarity values ​​of the above two knowledge block vectors belonging to the regulation text class to obtain the processed similarity value. At the same time, the knowledge block vectors (e.g., 3) matching the user information are determined based on multiple label information, and the second weight data corresponding to the user information is determined, and the second weight data is weighted for the similarity values ​​of the above three knowledge block vectors matching the user information to obtain the processed similarity value. Thus, the similarity values ​​of the 10 knowledge block vectors obtained after the above retrieval are updated based on the 5 processed similarity data. Among them, the label information of the knowledge block vector is obtained based on the material label data of the corresponding material data.

[0083] After obtaining the search results from the vector database, this embodiment further matches the label information of each knowledge block vector with the question category and user information corresponding to the query question, and weights the matched knowledge vectors to improve the similarity of the knowledge block vectors matching the question category in the search results and the similarity of the knowledge block vectors matching the user information, thereby obtaining updated search results and improving the accuracy of the similarity between each knowledge block vector in the search results and the query question, thereby avoiding the situation where the user obtains the operating procedures of a personal loan when querying about regulations related to a personal loan, further improving the accuracy and reliability of the request data fed back to the user's loan terminal, and thus reducing the loan delay.

[0084] S203. Obtain a target search result by sorting multiple knowledge block vectors in the updated search result, determine the request data based on the target search result, and feed it back to the user's loan terminal.

[0085] More specifically, the similarity between each knowledge block vector in the updated retrieval result and the query question is determined, and multiple knowledge block vectors are sorted based on the obtained similarities to obtain a target retrieval result; prompt data corresponding to the loan query request is generated based on the target retrieval result, and the corresponding request data is determined based on the prompt data and fed back to the user's loan terminal.

[0086] In a possible embodiment, based on the embodiment in step S202, the similarity values ​​corresponding to the 10 knowledge block vectors in the updated retrieval results are sorted in descending order, and a preset number of knowledge block vectors (such as 4) are selected from the sorted database vectors in descending order according to the similarity values, and corresponding prompt data are generated according to the 4 knowledge block vectors and the query question vector, and the request data corresponding to the query question vector is determined according to the prompt data and fed back to the user loan terminal.

[0087] The method for processing loan query requests provided in the embodiment of the present application obtains search results based on query questions, and further updates the search results based on user information and question categories, so as to obtain target search results by sorting the updated results again, thereby realizing automatic acquisition of request data corresponding to the query questions, improving the accuracy of the obtained request data, and improving loan efficiency.

[0088] In the prior art, information is usually delivered to account managers through offline training, email notifications, etc. However, due to the large number of employees of account managers, the information delivery method of the prior art is prone to cause information hysteresis effect, making it difficult to ensure that every employee can receive the latest information in a timely manner and apply it in business processing. As a result, there is a certain gap between the account manager's understanding of regulations and products and the latest status when handling business, which can easily mislead customers, thereby reducing the accuracy and efficiency of loan processing, and reducing the account manager's work experience and the customer's business processing experience.

[0089] Optionally, the material data set in the knowledge base also includes data to be revised uploaded by the review terminal.

[0090] Optionally, user query data is generated and sent to the audit terminal so that the audit terminal determines whether the user query data is data to be revised, and generates update data when it is determined to be data to be revised; the update data sent by the audit terminal is received, and the vector database is updated according to the update data.

[0091] In a possible embodiment, after the loan inquiry system feeds back the request data to the user's loan terminal, the loan inquiry system generates corresponding query data based on the question queried by the user and the obtained request data and sends it to the review terminal. The review terminal checks whether the request data corresponding to the loan inquiry request is accurate, and when it needs to be updated and corrected, uploads the data to be corrected (i.e., the material data after the request data is updated and corrected) to the knowledge base of the loan inquiry system, so that the loan inquiry system updates the vector database according to the material data updated in real time in the knowledge base. Among them, the update and correction operations include but are not limited to adding, deleting, modifying, and checking.

[0092] Optionally, the material data in the knowledge base is batch deleted every preset time period to timely clean up the expired and rigid material data in the knowledge base, release memory, and provide sufficient storage space for new material data.

[0093] This embodiment promptly reviews the request data fed back to the user through the audit terminal, and when it is determined that there is a lag or content error in the request data, the erroneous content is promptly corrected, the missing content is supplemented and adjusted, and uploaded to the knowledge base, so that the vector database is updated according to the corrected content through the loan query system, thereby improving the accuracy and stability of the vector database, and solving the technical problem that the existing technology is difficult to issue update information to a large number of account managers in a timely manner.

[0094] Optionally, historical loan query records are obtained from the user, and the historical loan query records include multiple historical loan query requests. When it is determined that there is an update of the knowledge block vector corresponding to any historical loan query request, historical update request data corresponding to the historical loan query request is generated based on the updated knowledge block vector and fed back to the user's loan terminal.

[0095] Exemplarily, the user loan terminal includes an auxiliary page and a business processing page. The auxiliary page is entered by clicking on the corresponding floating icon. The floating icon is a freely draggable icon covered on the business processing page. Therefore, this embodiment solves the problem of the prior art that problem query can be realized without constantly switching between pages through the floating icon, avoids frequent leaving the business processing page and affecting the business processing efficiency, improves the convenience of problem query, and further improves the efficiency of loan business processing.

[0096] Optionally, the interface of the auxiliary page includes a dialogue area and a function area, wherein the user inputs query questions through the dialogue area and obtains request data fed back by the loan query system through the dialogue area, and the function area also includes a common function area and a problem navigation area, and the common function area includes but is not limited to parameter configuration function, training function, pre-examination function, voice question and answer function and problem reporting function. Each of the above functions corresponds to a function icon presented in the corresponding function area of ​​the interface, and the corresponding function is entered by clicking the corresponding function icon. The problem navigation area is used to display historical loan query requests that users frequently search and collect. Users can obtain the corresponding request data by clicking any historical loan query request in the problem navigation area.

[0097] In a possible embodiment, when it is determined that the material data corresponding to any historical loan query request in the user's historical loan query record is corrected and updated, the corresponding updated knowledge block vector is determined in the vector database, and the update request data corresponding to the historical loan query request is generated according to the updated knowledge block vector, which is automatically sent to the user's loan terminal and presented in the training function area of ​​the user's loan terminal interface.

[0098] When the present embodiment determines that the knowledge block vector corresponding to any historical loan query request has been updated, it automatically generates updated request data based on the updated knowledge block vector and promptly presents it to the user's loan terminal interface, so that the user is promptly informed that the request data corresponding to the previously queried question has changed, and handles subsequent business based on the changed request data, thereby preventing the user from initiating the same query question again after querying a question because they are familiar with the request data corresponding to the question, and thus failing to handle subsequent business based on the changed relevant material data in a timely manner, resulting in errors in business handling. Therefore, the present embodiment improves the timeliness of the account manager's acquisition of information, avoids loan errors due to information lags, and improves the efficiency and reliability of loan business handling.

[0099] Optionally, real-time business information of the user's loan terminal is obtained; a target knowledge block vector is determined based on the real-time business information, and whether the real-time business information is valid is determined based on the target knowledge block vector; and when it is determined to be invalid, an alarm indication is generated based on the invalid business information and fed back to the user's loan terminal.

[0100] In a possible embodiment, when the query question is "application conditions for personal emergency loans", the loan query system obtains the real-time business information entered by the user in the user loan terminal interface in real time, and determines the target knowledge block vector that matches the filled-in item of the real-time business information in the vector database. When it is determined based on the text content of the target knowledge block vector that the real-time business information entered by the user exceeds the range indicated by the text content, the real-time business information is determined to be invalid, and the filled-in item is highlighted and fed back to the area corresponding to the pre-examination function of the user loan terminal interface to remind the user that the highlighted area currently filled in needs further review to see if there are any errors, so that the user can find errors in time, reduce the probability of human omissions, avoid certain losses to customers due to the issuance of an invalid application indication after a certain period of time after the application is submitted, reduce the time consumption of loan center staff due to low-level errors, and thus improve loan efficiency.

[0101] This embodiment monitors the business information handled by users in real time and simultaneously determines the validity of the business information, so as to promptly remind users when it is determined to be invalid or abnormal, so that users can take timely response measures to solve abnormal problems, reduce useless work, improve loan efficiency, and reduce losses to customers.

[0102] Optionally, the parameters in the loan inquiry system are configured through the parameter configuration function of the user loan terminal interface, for example, the preset quantity, so as to provide users with a loan inquiry system with parameter configuration that can meet user needs, improve the query efficiency of the loan inquiry system and the accuracy of the output request data, and enhance the user experience. The voice question and answer function of the user loan terminal interface is used to input the query questions in the dialogue area. The problem reporting function of the user loan terminal interface is used to propose suggestions for improving and optimizing the system, so that developers can optimize the system in a timely manner and improve the efficiency of business handling.

[0103] Optionally, the user loan terminal in the embodiment of the present application interacts with the loan inquiry system, wherein the user loan terminal can be a user mobile terminal or a business processing institution, and the embodiment of the present application does not limit this. When it is a user mobile terminal, the account manager can initiate a loan inquiry request to the loan inquiry system by text input or voice input in the dialogue area, so that the account manager can obtain the required request data in time when working outside, thereby improving the efficiency of the account manager's response to temporary emergency issues and the quality of business handling.

[0104] Figure 3 The structural diagram of the loan inquiry system provided for this application is as follows: Figure 3As shown, the loan query system includes a knowledge base, a preprocessing model, a vectorization model, a vector database, a retrieval model, a rearrangement model, a prompt word model, and a large language model. For example, the vectorization model is an Embedding model, the large language model is an LLM (Large Language Model) model, the rearrangement model is a Rerank model, and the prompt word model is a Prompt model including a Prompt template.

[0105] Optionally, the loan system includes a loan execution module and a user loan terminal. The user loan terminal sends a loan instruction to the loan execution module, so that the loan execution module executes the corresponding loan operation according to the loan instruction.

[0106] Optionally, when the user needs to query content during the business process, he / she can enter the dialogue area by clicking the floating icon on the user loan terminal interface and enter the query question in the dialogue area. The loan system calls the token interface based on the query question entered to obtain the token token to establish a dialog box. During the validity period of the token, the loan system sends the query question and user information entered by the user to the vectorized model of the loan query system.

[0107] In a possible embodiment, this embodiment Figure 3 On the basis of the embodiment, based on the loan inquiry system, a processing method for loan inquiry request is described in detail, which specifically includes the following steps: obtaining a material data set from a knowledge base through a preprocessing model, and performing data processing (such as filtering processing, compression processing, format processing) on ​​the material data set to obtain a plurality of text data of a target format corresponding to the material data set (each material data corresponds to a text data). By determining the text type of each text data, a segmentation strategy corresponding to the text type is obtained, and then data segmentation is performed on the above text data based on the obtained segmentation strategy to obtain a plurality of knowledge blocks corresponding to the text data, and input into a vectorization model. Each knowledge block is vectorized by the vectorization model, and the knowledge block vector corresponding to each knowledge block is obtained and stored in the vector database of the loan inquiry system.

[0108] In a possible embodiment, the query question is vectorized by the vectorization model in the loan query system to obtain the query question vector and input it into the retrieval model. The retrieval model uses the word vector similarity method to determine the retrieval range in the vector database and perform retrieval to obtain multiple similarities with the query question vector, and the knowledge block vector whose similarity exceeds the first preset similarity threshold is determined as the retrieval result. The retrieval model compares the label information corresponding to the multiple knowledge block vectors of the retrieval result according to the user information corresponding to the query question vector, obtains the knowledge block vector whose similarity with the user information exceeds the second preset similarity threshold, obtains the first weight data corresponding to the user information according to the weight mapping table, and performs weighted processing on the knowledge block vector whose similarity exceeds the second preset similarity threshold according to the first weight data to obtain the weighted knowledge block vector. At the same time, the retrieval model compares the label information corresponding to the multiple knowledge block vectors in the retrieval result according to the question category corresponding to the query question, and obtains the knowledge block vector whose similarity with the question category exceeds the third preset similarity threshold. According to the weight mapping table, the second weight data corresponding to the question category is obtained, and the knowledge block vector whose similarity exceeds the third preset similarity threshold is weighted according to the second weight data to obtain the weighted knowledge block vector. The updated retrieval result is obtained according to the weighted knowledge block vector.

[0109] In a possible embodiment, the updated retrieval results are sent to a rearrangement model, which calculates the similarity of each knowledge block vector with respect to the query question vector and reorders them in descending order of similarity. Based on the reordered results, a preset number of knowledge block vectors are selected as target retrieval results, and the target retrieval results and the query question vector are sent to a prompt word model.

[0110] In a possible embodiment, the prompt template in the prompt word model generates corresponding prompt information according to the target search result and the query question vector and sends it to the language model. The language model generates request data according to the prompt information and feeds it back to the user's loan terminal in the form of text or pictures.

[0111] Optionally, the language model sends the request data and the corresponding loan query request to the review terminal, so as to wait for receiving the correction instructions and correction content from the staff with review authority through the review terminal. When the review terminal receives the correction instructions and correction content, the correction content is transmitted to the knowledge base of the loan query system to realize timely correction of the request data and improve the real-time and accuracy of the request data.

[0112] Optionally, the revised content is also used to optimize and update the parameters of each model in the loan query system to improve the accuracy of the request data output by the loan query system. Parameter optimization and update include but are not limited to SFT (Supervised Fine-Tuning) full parameter tuning and Lora (Low-Rank Adaptation) fine tuning in the language model, as well as prompt tuning, rerank model tuning, and retrieval model tuning.

[0113] Exemplarily, the parameters of each model in the loan inquiry system are configured and updated according to user needs, wherein the user needs include but are not limited to the business processing institution and the business processing process.

[0114] Optionally, the user uploads different types of material data to the knowledge base through the user loan terminal, such as text, images, videos, etc. The loan query system analyzes and processes different types of data to automatically optimize the parameters of the loan query system. This embodiment improves the accuracy and robustness of the loan query system.

[0115] Optionally, the loan query system combines and processes text, image, video and other material data, and further improves the accuracy and robustness of the loan query system through the correlation and combined analysis between different types of data.

[0116] Optionally, the loan inquiry system can identify and process the emotional color and language style of various materials (such as official regulations, daily experience materials, and language expression habit materials), so as to obtain the corresponding emotions when the user sends a loan inquiry request, and accurately identify the user's request intention, thereby further improving the accuracy, intuitiveness and emotional richness of the request data and enhancing the user experience.

[0117] Optionally, when the user obtains the request data in the dialogue area of ​​the user's loan terminal, the user can interact with the dialogue area and input evaluation information on the query result of this problem to the loan query system. The loan query system optimizes and updates the parameters of each model in the loan query system according to the evaluation information, further enhancing the query accuracy and reliability of the loan query system. The input form of the evaluation information includes but is not limited to likes, dislikes, and problem reporting.

[0118] In a possible embodiment, the loan system also includes a data display module, which displays the number of queries, user satisfaction, user click request data and reported problem data of the same user within a fixed period in the form of icons, so that developers can analyze and optimize the loan query system and user loan terminals. For example, based on the query questions that users frequently like, the request data of the query question is pushed to the front of the question navigation area of ​​the user loan terminal to facilitate user query.

[0119] In a possible embodiment, the loan system sends the data of user reported problems and high-frequency step-down data to the audit terminal, so that the audit terminal can correct the above data to improve the accuracy of the knowledge base material data and the timeliness of the update.

[0120] Optionally, the language big model is also used to obtain real-time business data of the user's loan terminal, and verify it according to the regulations, operating manuals, business expert experience and system rules updated in real time in the knowledge base to determine whether the current business processing page contains obvious errors, does not comply with system rules, and does not comply with expert experience data. If it is determined that there are such errors, the content will be highlighted and fed back to the interface of the user's loan terminal to remind the user to check again.

[0121] While feeding back accurate request data to the user's loan terminal, this embodiment will highlight and remind the user the filled-in content that obviously does not meet the management requirements, experience data or system rules based on the material data updated in real time in the knowledge base, so as to facilitate the user to find errors in time, reduce the probability of human omissions, reduce the time consumption of loan staff due to low-level errors, and improve loan efficiency.

[0122] Figure 4 A schematic diagram of the structure of the processing device for the loan inquiry request provided by this application, such as Figure 4 As shown, the processing device 40 for loan inquiry request provided in this embodiment is applied to the loan inquiry system and includes:

[0123] The acquisition module 401 is used to acquire a loan query request sent by a user's loan terminal, process the loan query request, and obtain a query question vector. The loan query request includes a query question and user information.

[0124] Processing module 402, for determining a search result corresponding to the query question based on the vector database, the search result including a plurality of knowledge block vectors; and obtaining an updated search result according to the user information and the question category corresponding to the query question;

[0125] Processing module 402 is also used to determine the similarity between each knowledge block vector in the updated search results and the query question, and sort multiple knowledge block vectors based on the obtained similarities to obtain a target search result; generate prompt data corresponding to the loan query request based on the target search result, determine the corresponding request data based on the prompt data and feed it back to the user loan terminal.

[0126] Optionally, the acquisition module 401 is further used to acquire a material data set updated in real time before determining the search result, where the material data set includes a plurality of material data and material tag information corresponding to each material data;

[0127] By preprocessing the material data set, a plurality of text data in a target format are obtained, and the text type of each text data is determined;

[0128] For any text data type, the corresponding segmentation strategy is determined based on the text type, and the text data is segmented based on the segmentation strategy to obtain multiple text knowledge blocks corresponding to the text data. The multiple text knowledge blocks are vectorized to obtain multiple knowledge block vectors corresponding to the text data and stored in the vector database.

[0129] Optionally, the processing module 402 is further configured to obtain updated search results according to the user information and the question category corresponding to the query question, specifically including:

[0130] Determine label information of each knowledge block vector in the retrieval result based on the material label information;

[0131] Determine the similarity between each tag information in the search result and the user information to obtain a first knowledge block vector set, and determine the similarity between each tag information and the question category to obtain a second knowledge block vector set;

[0132] Based on the weight mapping table, determine the first weight data corresponding to the user information and the second weight data corresponding to the question category; perform weighted processing on the first knowledge block vector set according to the first weight data, and perform weighted processing on the second knowledge block vector set according to the second weight data;

[0133] The updated retrieval result is determined based on the weighted knowledge block vector.

[0134] Optionally, the acquisition module 401 is further used to obtain real-time business information of the user's loan terminal;

[0135] A target knowledge block vector is determined based on real-time business information, and whether the real-time business information is valid is determined based on the target knowledge block vector; and when it is determined to be invalid, an alarm indication is generated based on the invalid business information and fed back to the user's loan terminal.

[0136] Optionally, the processing module 402 is further used to generate user query data and send it to the audit terminal, so that the audit terminal determines whether the user query data is data to be revised, and generates update data when it is determined to be data to be revised;

[0137] Receive the update data sent by the audit terminal, and update the vector database according to the update data.

[0138] Optionally, the acquisition module 401 is further used to acquire a historical loan query record of the user, where the historical loan query record includes multiple historical loan query requests;

[0139] When it is determined that there is an update of the knowledge block vector corresponding to any historical loan query request, historical update request data corresponding to the historical loan query request is generated based on the updated knowledge block vector and fed back to the user loan terminal.

[0140] The processing device for loan inquiry request provided in this embodiment is applied to the loan inquiry system and can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar and will not be described in detail in this embodiment.

[0141] Figure 5 This is a schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device 50 provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the device 50 also includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus 504.

[0142] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that at least one processor 501 executes the above method.

[0143] The specific implementation process of the processor 501 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0144] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0145] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0146] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0147] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0148] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0149] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0150] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0151] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0152] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0153] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0154] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0155] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0156] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for processing a loan inquiry request, characterized in that: Applied to loan inquiry system, including: Acquire a loan query request sent by a user loan terminal, process the loan query request, and obtain a query question vector, wherein the loan query request includes a query question and user information; Determine a search result corresponding to the query question based on a vector database, wherein the search result includes a plurality of knowledge block vectors; obtain an updated search result according to the user information and the question category corresponding to the query question; Determine the similarity between each knowledge block vector in the updated search result and the query question, and sort the multiple knowledge block vectors based on the obtained similarities to obtain a target search result; generate prompt data corresponding to the loan query request based on the target search result, determine the corresponding request data based on the prompt data and feed it back to the user loan terminal.

2. The method according to claim 1, characterized in that Also includes: Before determining the search result, obtaining a material data set updated in real time, the material data set including a plurality of material data and material tag information corresponding to each material data; By preprocessing the material data set, a plurality of text data in a target format are obtained, and the text type of each text data is determined; For any text data type, a corresponding segmentation strategy is determined based on the text type, and the text data is segmented based on the segmentation strategy to obtain multiple text knowledge blocks corresponding to the text data, and the multiple text knowledge blocks are vectorized to obtain multiple knowledge block vectors corresponding to the text data and store them in a vector database.

3. The method according to claim 2, characterized in that According to the user information and the question category corresponding to the query question, an updated search result is obtained, specifically including: Determine label information of each knowledge block vector in the retrieval result based on the material label information; Determine the similarity between each tag information in the search result and the user information to obtain a first knowledge block vector set, and determine the similarity between each tag information and the question category to obtain a second knowledge block vector set; Based on the weight mapping table, determine the first weight data corresponding to the user information and the second weight data corresponding to the question category; perform weighted processing on the first knowledge block vector set according to the first weight data, and perform weighted processing on the second knowledge block vector set according to the second weight data; The updated retrieval result is determined based on the weighted knowledge block vector.

4. The method according to any one of claims 1 to 3, characterized in that: Also includes: Obtaining real-time business information of the user's loan terminal; A target knowledge block vector is determined based on the real-time business information, and whether the real-time business information is valid is determined based on the target knowledge block vector; and when it is determined to be invalid, an alarm indication is generated based on the invalid business information and fed back to the user's loan terminal.

5. The method according to claim 4, characterized in that Also includes: Generate user query data and send it to the audit terminal, so that the audit terminal determines whether the user query data is data to be revised, and generates update data when it is determined to be data to be revised; The update data sent by the audit terminal is received, and the vector database is updated according to the update data.

6. The method according to claim 5, characterized in that Also includes: Acquire a historical loan query record from a user, wherein the historical loan query record includes a plurality of historical loan query requests; When it is determined that there is an update of the knowledge block vector corresponding to any historical loan query request, historical update request data corresponding to the historical loan query request is generated based on the updated knowledge block vector and fed back to the user loan terminal.

7. A processing device for loan inquiry request, characterized in that: Applied to loan inquiry system, including: An acquisition module, used to acquire a loan query request sent by a user's loan terminal, process the loan query request, and obtain a query question vector, wherein the loan query request includes a query question and user information; A processing module, configured to determine a search result corresponding to the query question based on a vector database, wherein the search result includes a plurality of knowledge block vectors; and obtain an updated search result according to the user information and the question category corresponding to the query question; The processing module is also used to determine the similarity between each knowledge block vector in the updated retrieval result and the query question, and sort the multiple knowledge block vectors based on the obtained similarities to obtain a target retrieval result; generate prompt data corresponding to the loan query request based on the target retrieval result, determine the corresponding request data based on the prompt data and feed it back to the user loan terminal.

8. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images