Credited standardized reply text generation method and device, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence and financial science and technology, and discloses a standardized reply text generation method and device for credit granting, equipment and a medium, and the method comprises the steps: obtaining a historical reply statement of a credit granting product, carrying out the label division of the historical reply statement, and obtaining a reply label corresponding to the historical reply statement; performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors; clustering the reply category vectors to obtain a reply category set, selecting representative statements in the reply category set, and generating standard reply suggestions according to the representative statements; obtaining product parameters of the target product, and adding the standard reply opinions to a pre-constructed reply template according to the product parameters; and generating a standardized reply text of the target product according to the reply template added with the standard reply opinions. According to the invention, the efficiency of reply opinion feedback can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and financial technology, and in particular to a method, device, equipment and medium for generating a standardized approval text for credit. Background Art

[0002] Among the corporate business of banks, corporate loans are the main profitable products of banks. When an enterprise applies for a corporate loan, it needs to go through the credit approval of the bank's loan system, sign a contract, make a loan, and monitor the loan after it is loaned, and the enterprise finally gets the loan. In the credit approval process, the account manager who helps the enterprise to apply for the loan needs to investigate the enterprise, and the approving account manager issues a credit approval reply. The post-loan account manager who disburses the loan needs to implement the loan conditions according to the reply, so the credit approval reply is very critical in the entire loan process.

[0003] In the existing process, approval opinions are divided into pre-loan restrictive opinions, opinions on signing contract terms, and post-loan restrictive conditions opinions. Because there are a large number of account managers who approve loans, the credit approval opinions are very personalized and are not written according to specific modules. For example, the approval opinions involving post-loan monitoring are written into the pre-loan restrictive opinions, causing the account managers responsible for subsequent loan issuance to repeatedly communicate the post-loan opinions, resulting in high communication costs and problems with loan monitoring, which leads to low efficiency in feedback on approval opinions. Summary of the invention

[0004] The present invention provides a method, device, equipment and medium for generating a standardized approval text for credit, so as to solve the technical problem of low efficiency in providing feedback on approval opinions.

[0005] In a first aspect, a method for generating a standardized approval text for credit is provided, comprising:

[0006] Obtain historical approval statements of the credit product, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements;

[0007] Performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors;

[0008] Clustering the reply category vectors to obtain a reply category set, selecting representative sentences from the reply category set, and generating standard reply opinions based on the representative sentences;

[0009] Obtain product parameters of the target product, and add the standard approval opinion to a pre-built approval template according to the product parameters;

[0010] Generate a standardized approval text for the target product based on the approval template after adding standard approval opinions.

[0011] In a second aspect, a device for generating a standardized approval text for credit is provided, comprising:

[0012] An approval label classification module is used to obtain historical approval statements of credit products, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements;

[0013] A reply category sentence classification module is used to perform text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and convert the reply category sentences into reply category vectors;

[0014] A standard reply opinion generating module is used to cluster the reply category vectors to obtain a reply category set, select representative sentences from the reply category set, and generate a standard reply opinion based on the representative sentences;

[0015] An approval template filling module is used to obtain product parameters of a target product and add the standard approval opinions to a pre-built approval template according to the product parameters;

[0016] The standardized approval text generation module is used to generate the standardized approval text of the target product according to the approval template after adding the standard approval opinions.

[0017] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for generating a standardized approval text for credit are implemented.

[0018] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the method for generating a standardized approval text for credit are implemented.

[0019] The embodiment of the present invention extracts some samples from the historical approval opinions and marks them, and divides the approval opinions into three module statements such as pre-loan, contract, and post-loan through supervised text classification, and then performs text clustering on the opinions in each module to obtain the clustered text, and uses a similarity algorithm to compare the texts in pairs to obtain the representative text in each class, thereby obtaining the pre-loan, contract, and post-loan standard statements, and can automatically generate the corresponding approval according to the historical approval of the account manager and the credit product parameters with one key, and the approval terms can be accurately filled in the corresponding pre-loan and post-loan modules. The approval opinions are highly standardized, which can effectively improve the communication efficiency between the loan center and the approval center, and the loan account manager can perform the corresponding operations according to the corresponding terms. The post-loan account manager can strengthen the monitoring of the loan, discover loan anomalies in advance, and reduce the default rate of the loan. Therefore, the standardized approval text generation method, device, equipment and medium for credit proposed by the present invention can solve the problem of low efficiency when giving feedback on approval opinions. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0021] Figure 1 It is a schematic diagram of an application environment of a method for generating a standardized approval text for credit granting in one embodiment of the present invention;

[0022] Figure 2 It is a flow chart of a method for generating a standardized approval text of credit in one embodiment of the present invention;

[0023] Figure 3 yes Figure 1 A schematic flow chart of a specific implementation method of step S2;

[0024] Figure 4 yes Figure 1 A schematic flow chart of a specific implementation of step S3;

[0025] Figure 5 It is a structural schematic diagram of a device for generating a standardized reply text for credit granting in one embodiment of the present invention;

[0026] Figure 6 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0027] Figure 7 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The method for generating a standardized credit approval text provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In an application environment, the client communicates with the server through a network. The server can obtain historical approval statements of credit products through the client, label the historical approval statements, and obtain approval labels corresponding to the historical approval statements; perform text classification on the historical approval statements according to the approval labels to obtain approval category statements, and convert the approval category statements into approval category vectors; cluster the approval category vectors to obtain an approval category set, select representative statements from the approval category set, and generate standard approval opinions based on the representative statements; obtain product parameters of the target product, and add the standard approval opinions to a pre-built approval template based on the product parameters; generate a standardized approval text for the target product based on the approval template after adding the standard approval opinions. In the present invention, by extracting some samples from the historical approval opinions and marking them, the approval opinions are divided into three module statements such as pre-loan, contract, and post-loan through supervised text classification, and then the opinions in each module are subjected to text clustering to obtain the clustered text. The texts are compared in pairs using a similarity algorithm to obtain representative texts in each class, thereby obtaining pre-loan, contract, and post-loan standard statements. The corresponding approval can be automatically generated according to the historical approval of the account manager and the credit product parameters with one key. The approval terms can be accurately filled in the corresponding pre-loan and post-loan modules. The approval opinions are highly standardized, which can effectively improve the communication efficiency between the loan center and the approval center. The loan account manager can perform corresponding operations according to the corresponding terms. The post-loan account manager can strengthen the monitoring of the loan, discover loan anomalies in advance, and reduce the default rate of the loan. Among them, the client can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server can be implemented with an independent server or a server cluster composed of multiple servers. The present invention is described in detail below through specific embodiments.

[0030] See also Figure 2 As shown, Figure 2 A flow chart of a method for generating a standardized approval text for credit provided by an embodiment of the present invention includes the following steps:

[0031] S1. Obtain historical approval statements of a credit product, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements.

[0032] In an embodiment of the present invention, the historical approval statements refer to those written by the account manager and the loan reviewer, such as before the loan: complete the joint and several liability guarantee procedures of xx company and xx individual; this credit is included in the credit management of xx group; contract: the mortgaged property is evaluated every year, and the loan mortgage rate shall not exceed xx; after the loan: the branch / operation team establishes a lease contract registration book and a detailed ledger of lease status, records the property rental status and rent payment status in a timely manner, accurately estimates the operating cash flow, and dynamically adjusts the repayment plan.

[0033] In detail, historical approval statements of credit products can be obtained from a pre-stored storage area through computer statements with data capture functions (such as Java statements, Python statements, etc.), where the storage area includes but is not limited to databases and blockchains.

[0034] Furthermore, in order to classify all historical reply statements into different categories and facilitate the subsequent generation of standardized reply texts, it is necessary to perform labeling operations on the historical reply statements to obtain reply statements of different categories.

[0035] In the embodiment of the present invention, the approval labels refer to three labels: pre-loan, contract, and post-loan.

[0036] In the embodiment of the present invention, the labeling of the historical reply statements to obtain the reply labels corresponding to the historical reply statements includes:

[0037] Selecting a target reply statement from the historical reply statements by a preset random sampling;

[0038] Segment the target reply sentence to obtain a target reply segmented sentence;

[0039] Labeling the target approval segmentation sentence according to a preset loan label to obtain a sentence label corresponding to the target approval segmentation sentence;

[0040] Training a preset classification model through the target approval segmentation sentences and the sentence labels;

[0041] The trained classification model is used to perform label classification on the historical reply statements to obtain the reply labels corresponding to the historical reply statements.

[0042] Specifically, target approval statements in historical approval statements are obtained by random sampling according to credit products, that is, the data set including historical approval statements is sorted to ensure that the data contains relevant information of credit products, and corresponding credit product labels are added to each historical approval statement, such as "personal loan", "corporate loan", "credit card", etc. The historical approval statements are classified according to credit products to form multiple subsets, each subset corresponds to a credit product, the sample size of each credit product is determined according to random sampling, and random sampling is performed on the subset of each credit product to ensure that the sample is representative. Simple random sampling or stratified sampling can be used to extract the target approval statement from each sampling result to form a final sample set to ensure that the selected target approval statement is representative.

[0043] Specifically, natural language processing (NLP) tools are used to perform syntactic analysis and segmentation, the target approval sentence is split into smaller sentence units, and the entire target approval sentence is segmented by periods to obtain the segmented target approval segmented sentences. Then, a small amount of manual classification and labeling is performed to manually annotate each segmented sentence, and the pre-loan, contract, and post-loan labels are manually labeled. Then, the annotated target approval segmented sentences and their corresponding sentence labels are trained through the TextCNN classification method. That is, the text is segmented through Tokenizer, and the text is converted into a sequence, the text label is encoded into an integer, and to_categorical is used for One-hot encoding is used to define the TextCNN network structure, including an embedding layer, multiple convolutional layers and a pooling layer. The training set is used for training, and the model performance is evaluated on the validation set. The hyperparameters (such as the maximum number of words, the number of samples, the number of training times, etc.) are continuously adjusted to obtain the best results. The trained classification model is then used to classify the labels of historical approval statements, and all historical approval statements are divided into three categories: pre-loan, contractual, and post-loan. That is, the historical approval statements are taken as feature vectors, and the feature vectors are input into the trained classification model. The output index is converted into an actual label, namely, pre-loan, contractual, and post-loan approval labels, thereby achieving labeling and classification of all historical statements.

[0044] Furthermore, in order to generate different approval opinions, it is necessary to classify historical approval statements so as to obtain approval opinions under different categories, thereby effectively improving the communication efficiency between the loan center and the approval center.

[0045] S2. Performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors.

[0046] In an embodiment of the present invention, the approval category statement refers to the specific statement in each category obtained after classifying the historical approval statements according to specific labels, that is, the specific approval statements corresponding to the three label categories of pre-loan, contract, and post-loan are obtained.

[0047] In the embodiment of the present invention, referring to Figure 3 As shown, the text classification of the historical reply statements is performed according to the reply labels to obtain reply category statements, including:

[0048] S31, matching the reply label with the historical reply sentence to obtain a sentence label mapping relationship;

[0049] S32, dividing the historical approval statements into pre-loan approval statements, contract approval statements and post-loan approval statements according to the statement label mapping relationship;

[0050] S33, determining the pre-loan approval statement, the contract approval statement and the post-loan approval statement as the approval category statements.

[0051] In detail, the approval label refers to the pre-loan label, the contract label and the post-loan label. The pre-loan approval includes the loan application review, the amount annotation and other related contents; the contract approval includes the confirmation, modification and signing of the loan contract; the post-loan approval includes the management after the use of the loan, the repayment arrangement and other follow-up matters, and then the historical approval statements are matched according to the approval label, that is, the pre-loan label is matched with the pre-loan approval statement, the contract label is matched with the contract approval statement, and the post-loan label is matched with the post-loan approval statement. If the approval label is a pre-loan approval, the corresponding pre-loan approval statement is "Agree to your loan application", "The loan application is not approved, and additional materials are required"; if the approval label is a contract approval, the corresponding contract approval statement is "Please sign the loan contract and submit", "The contract content has been modified, please confirm"; if the approval label is a post-loan approval, the corresponding post-loan approval statement is "Your loan has been successfully issued, please repay on time", "Please repay at the agreed time, and interest will be charged for overdue payment".

[0052] Specifically, the statements are classified according to the statement label mapping relationship, and the pre-loan approval statements are "Agree with your loan application", "The loan application is not approved, and additional materials are required"; the contract approval statements are "Please sign the loan contract and submit", "The contract content has been modified, please confirm"; the post-loan approval statements are "Your loan has been successfully issued, please repay on time", "Please repay on the agreed time, and interest will be charged for overdue repayment". Therefore, the final classification results of the historical approval statements are pre-loan approval statements, contract approval statements and post-loan approval statements.

[0053] Furthermore, among the classified pre-loan approval statements, contract approval statements and post-loan approval statements, there are many statements that differ by only a few words, or are sub-clauses of a certain sentence. For example, completing the joint and several liability guarantee procedures of xx company, completing the joint and several liability guarantee procedures of xx individual and completing the joint and several liability guarantee procedures of xx company and xx individual are very similar. In this case, the approval category statements need to be clustered to obtain more refined approval category statements.

[0054] In an embodiment of the present invention, before clustering the reply category sentences, it is necessary to convert the reply category sentences into reply category vectors to prepare for subsequent clustering analysis, wherein the reply category sentences can be converted into reply category vectors through a vector conversion model, and the vector conversion model includes but is not limited to a BERT model and a Word2Vec model. The BERT model is used to generate context-related sentence vectors, and then a clustering algorithm (such as K-means, DBSCAN, etc.) is used to perform cluster analysis on the generated vectors to identify similar reply category sentences.

[0055] Furthermore, in order to identify similar approval categories and thus reveal the relationships and patterns between different categories, it is necessary to cluster similar approval categories together, which will help identify and eliminate duplicate or redundant approval categories and improve the integration and consistency of the data.

[0056] S3. Clustering the reply category vectors to obtain a reply category set, selecting representative sentences from the reply category set, and generating standard reply opinions based on the representative sentences.

[0057] In an embodiment of the present invention, the approval category set refers to a group of categories formed by classifying and summarizing various approval documents or information after clustering analysis of similar statements in the three categories of pre-loan, contract, and post-loan, and each category represents approval content with similar characteristics or themes.

[0058] In detail, the sentence vectors are clustered by kmeans clustering to obtain classification. First, the number of cluster categories K is determined, such as by plotting the sum of squared clustering errors (SSE) under different K values, finding the inflection point of the "S-shaped" graph or evaluating the ratio of the average distance of a data point within its own category to the average distance to its nearest category. The higher the value, the better the clustering effect. K initial centroids are randomly selected, and each sentence vector is assigned to the category where the nearest centroid is located. The centroid is recalculated based on all sentence vectors in the current category, that is, the mean of all points in each category is calculated. The assignment and update process is repeated until the centroid no longer changes significantly or reaches the preset number of iterations. Then, the sentence is associated with its corresponding cluster label to understand the characteristics and representative sentences of each category. Analyzing the clustering results can help identify the similarities and topics between sentences, where K-means relies on a distance metric (usually Euclidean distance). By constantly adjusting the centroid, K-means attempts to minimize the variability of samples within each cluster (i.e., SSE).

[0059] Furthermore, the clustered sentences are compared for similarity to obtain representative sentences in each clustering result. That is, after completing K-means clustering, each sentence is assigned to a specific cluster, and the most representative sentence needs to be selected from each cluster to obtain standard sentences before loan, contract, and after loan.

[0060] In the embodiment of the present invention, the representative statement refers to the statement that can most effectively summarize, reflect or represent the overall meaning or characteristics of the group of statements among the three types of statements: pre-loan, contractual, and post-loan. It has a high similarity with other statements and shows consistency.

[0061] In the embodiment of the present invention, the selecting of representative sentences in the approval category set includes:

[0062] Extracting all reply statements in the reply category set;

[0063] Calculate the similarity between adjacent reply statements;

[0064] The reply sentence with the highest similarity is selected as the representative sentence.

[0065] In detail, it is implemented through text processing tools or programming languages ​​(such as Python) to ensure that each approval statement is listed separately, that is, from the approval category set, first extract all the approval statements, convert each approval statement into a vector, and use the cosine similarity formula to calculate the similarity between the statements. For example, in Python, the cosine_similarity function in the sklearn library can be used to calculate the cosine similarity, and the similarity calculation results between all the approval statements are organized into a matrix. For each approval statement, find the other statements with the highest similarity to it, and use them as representative statements, that is, according to the similarity score, select the top N most similar sentences from the cluster as representative statements. The representative statements can summarize the main content or theme of the cluster and help understand the characteristics of different categories.

[0066] For example, a representative statement before a loan is "Please submit your income certificate and credit report for our approval"; a representative statement in a contract is "This contract will take effect from the date of signature by both parties, and the loan amount and interest rate will be specified in the contract;" a representative statement after a loan is "Please repay on time to avoid late fees and affect your credit record."

[0067] Furthermore, based on representative statements, it can ensure that the replies given by different personnel when dealing with similar situations remain consistent, reducing the influence of subjective factors, so that standardized statements can be generated based on representative statements. Standardized statements can speed up the approval process, and staff can quickly select appropriate replies without having to write from scratch each time, thereby improving work efficiency.

[0068] In an embodiment of the present invention, the standard reply opinion refers to a unified response template formulated for pre-loan, contractual, and post-loan requests. The standard reply opinion contains clear, concise, and professional language, aiming to provide consistency and standardization to improve communication efficiency. The standard reply opinion is typically used in areas such as customer service, administrative approval, and loan management to ensure that all relevant personnel follow the same format and tone when expressing similar content.

[0069] In the embodiment of the present invention, referring to Figure 4 As shown, the generating of standard reply opinions according to the representative sentences includes:

[0070] S41, extracting the reply template corresponding to the representative sentence;

[0071] S42, performing semantic analysis on the representative sentences to obtain the emotional semantics and content semantics corresponding to the representative sentences;

[0072] S43, generating the template content of the reply template according to the emotional semantics and the content semantics;

[0073] S44. Generate the standard reply opinion according to the template content.

[0074] In detail, the approval template refers to the standardized response format used when processing specific transactions or requests, including preset content, structure and language, which aims to provide fast, consistent and professional responses. For example, the representative statement before the loan is "Please provide your personal information and financial situation for preliminary assessment", and its corresponding approval template is "Thank you for your application. For further processing, please provide your personal information and financial situation"; the representative statement of the contract is "This loan contract contains terms such as the loan amount, interest rate and repayment period", and its corresponding approval template is "Your loan contract has been prepared, which lists the terms such as the loan amount, interest rate and repayment period in detail"; the representative statement after the loan is "Please check your account balance regularly and ensure repayment on time", and its corresponding approval template is "Please check your account balance regularly and repay on time to avoid additional fees."

[0075] Specifically, natural language processing (NPL) is used to perform semantic analysis on representative sentences to obtain the emotional semantics and content semantics corresponding to the representative sentences. The emotional semantics before the loan is neutral (requesting information), and the content semantics is to collect customer information for evaluation; the emotional semantics of the contract is positive (notifying the customer of the contract preparation status), and providing contract details; the emotional semantics after the loan is neutral (reminders and suggestions), and the content semantics is to remind the customer to pay attention to the account and repayment, and then generate the content of the approval template based on the emotional semantics and content semantics. That is, the content of the approval template refers to the information and structure contained in the template, which is usually used to respond to requests or transaction processing, including salutations, thank you words, confirmation of applications / requests, processing results, follow-up steps, contact information, and closing remarks. The content that needs to be filled in the approval template is determined according to the emotional semantics and content semantics, and then a standard approval opinion is generated according to the template content. For example, the standard approval opinion before the loan is Dear customer, thank you for submitting your loan application. In order to better serve you, please provide the necessary personal information and financial situation, and we will evaluate you as soon as possible; the standard reply to the contract is Dear customer, we are pleased to inform you that your loan contract is ready, which details the loan amount, interest rate and repayment period. If you have any questions, please feel free to contact us; the standard reply after the loan is Dear customer, in order to ensure the smooth progress of your loan, please check the account balance regularly and make sure to repay on time to avoid unnecessary expenses. If you have any questions, please feel free to contact our customer service.

[0076] Furthermore, standardized approval statements can be obtained from historical approval statements through the model, and then automatically filled in the corresponding modules according to product parameters, standardized statements can be used to reduce communication costs and improve work efficiency.

[0077] S4. Obtain product parameters of the target product, and add the standard approval opinion to a pre-built approval template according to the product parameters.

[0078] In an embodiment of the present invention, the product parameters include the loan product name, loan term, and loan amount, wherein the product parameters of the target product can be obtained from a pre-stored storage area through a computer statement with a data capture function (such as a Java statement, a Python statement, etc.), and the storage area includes but is not limited to a database and a blockchain.

[0079] Furthermore, by combining specific product parameters, the approval content can be made more targeted and professional, reflecting the importance attached to customer needs. Detailed product parameters can make the approval information clearer, helping customers understand relevant decisions or requirements, thereby reducing misunderstandings and the complexity of subsequent communication.

[0080] In the embodiment of the present invention, when the system is conducting approval, the product parameters are passed to the approval opinion function, and the approval statement of the corresponding module can be returned, and the corresponding module is automatically filled in the front end.

[0081] In the embodiment of the present invention, the adding the standard reply opinion to the pre-built reply template according to the product parameters includes:

[0082] Divide the pre-built approval templates into rejection approval templates and approval approval templates according to preset customer needs;

[0083] When the customer demand is an approval demand, the product approval opinions corresponding to the product parameters are selected from the standard approval opinions, and the product approval opinions are added to the approval template;

[0084] When the customer demand is a rejection demand, the customer manager's feedback rejection reason is extracted and the feedback rejection reason is added to the rejection reply template.

[0085] In detail, the rejection reply template is specifically used to handle situations where customer needs are rejected, and the approval reply template is used to confirm and agree to customer needs. Furthermore, when the customer demand is an approval demand, product approval opinions corresponding to the product parameters are screened out from the pre-built standard approval opinions based on the customer's specific needs, and the screened product approval opinions are integrated into the approval reply template to ensure the integrity and consistency of the content; when the customer demand is a rejection demand, the rejection reason is extracted from the customer manager's feedback, and the extracted feedback rejection reason is integrated into the rejection reply template so as to explain the reason for rejection in detail to the customer. Through this structured method, it is possible to more efficiently generate approval opinions that meet different customer needs and improve response speed and accuracy.

[0086] For example, the product parameters are input: loan product name, loan term, loan amount, and the approving account manager clicks "Agree". The product parameters are passed from the front end to the back end model to generate a reply opinion. If "Reject" is clicked, a rejection reply template is generated and the account manager needs to manually fill in the reason for rejection. The pre-loan restriction opinion, contract signing opinion, and post-loan restriction opinion returned by the model are automatically filled in the corresponding module. The account manager can add or subtract opinions from a certain module based on actual conditions.

[0087] Furthermore, by standardizing the approval opinions, each template can be automatically filled in with the correct approval opinion statements, which can improve efficiency and reduce risks for the company. When customer needs change, the template content can be quickly adjusted to ensure that the approval still meets the latest requirements.

[0088] S5. Generate a standardized approval text for the target product based on the approval template after adding standard approval opinions.

[0089] In the embodiment of the present invention, the standardized reply text refers to a unified and standardized written reply generated according to a pre-set format and content template for a specific situation or need.

[0090] In the embodiment of the present invention, the step of generating a standardized reply text for a target product according to the reply template after adding the standard reply opinion includes:

[0091] Extract the template format of the reply template after adding standard reply opinions;

[0092] Converting the template format into a text format;

[0093] The approval template after adding the standard approval opinions is used to generate a standardized approval text for the target product according to the text format.

[0094] In detail, extract the template format corresponding to the reply template after adding the standard reply opinion. For example, the reply template format is Dear [Customer Name], thank you for your attention to our products. In response to the [specific needs or questions] you raised, we have carefully evaluated and now give our reply opinions as follows: 1. **Product Name**: [Target Product Name], 2. **Reply Opinion**: -[Standard Reply Opinion 1], -[Standard Reply Opinion 2], -[Other relevant information], If you have any questions, please feel free to contact us. Thank you for your understanding and support! Sincerely, [Company Name], [Date], and convert the template format to text format, then the reply template in text format is Dear XX, thank you for your attention to our products. In response to your technical support request, we have carefully evaluated and now give our reply opinions as follows: 1. **Product Name**: XYZ Smart Device, 2. **Reply Opinion**: -We confirm that the product meets your technical specifications. -It is recommended to upgrade the system before use to ensure optimal performance. If you have any questions, please feel free to contact us. Thank you for your understanding and support! Sincerely, XX Technology Co., Ltd., XX / XX / XX, thereby generating a standardized approval text for the target product.

[0095] For example, the standardized reply text is Dear XX, thank you for your attention to our products. After careful evaluation of your technical support request, we now provide our reply as follows: 1. **Product Name**: XYZ Smart Device, 2. **Reply**: - We confirm that the product meets your technical specifications. - It is recommended to upgrade the system before use to ensure optimal performance. If you have any questions, please feel free to contact us. Thank you for your understanding and support! Sincerely, XX Technology Co., Ltd., XX / XX / XX, thereby generating a standardized reply text for the target product, so that the standardized reply text for the target product can be efficiently generated to ensure the clarity and professionalism of information communication.

[0096] It can be seen that in the above scheme, by extracting some samples from the historical approval opinions and marking them, the approval opinions are divided into three module statements such as pre-loan, contract, and post-loan through supervised text classification, and then the opinions in each module are clustered to obtain the clustered text. Using the similarity algorithm, the texts are compared two by two to obtain the representative text in each class, thereby obtaining the pre-loan, contract, and post-loan standard statements. According to the historical approval of the account manager and the parameters of the credit product, the corresponding approval can be automatically generated with one key, and the approval terms can be accurately filled in the corresponding pre-loan and post-loan modules. The approval opinions are highly standardized, which can effectively improve the communication efficiency between the loan center and the approval center. The loan account manager can perform the corresponding operations according to the corresponding terms, and the post-loan account manager can strengthen the monitoring of the loan, discover loan anomalies in advance, and reduce the default rate of the loan. Therefore, the standardized approval text generation method, device, equipment and medium of the credit proposed in the present invention can solve the problem of low efficiency when giving feedback on approval opinions.

[0097] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0098] In one embodiment, a device for generating a standardized reply text for credit is provided, and the device for generating a standardized reply text for credit corresponds to the method for generating a standardized reply text for credit in the above embodiment. Figure 5 As shown, the standardized reply text generation device for credit granting includes a reply label classification module 101, a reply category statement classification module 102, a standard reply opinion generation module 103, a reply template filling module 104 and a standardized reply text generation module 105. The detailed description of each functional module is as follows:

[0099] The approval label classification module 101 is used to obtain historical approval statements of credit products, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements;

[0100] The reply category sentence classification module 102 is used to perform text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and convert the reply category sentences into reply category vectors;

[0101] The standard reply opinion generating module 103 is used to cluster the reply category vectors to obtain a reply category set, select representative sentences in the reply category set, and generate a standard reply opinion according to the representative sentences;

[0102] The approval template filling module 104 is used to obtain product parameters of the target product and add the standard approval opinions to the pre-built approval template according to the product parameters;

[0103] The standardized reply text generating module 105 is used to generate a standardized reply text of a target product according to the reply template after adding standard reply opinions.

[0104] In one embodiment, the approval label classification module 101 is specifically used to:

[0105] Selecting a target reply statement from the historical reply statements by a preset random sampling;

[0106] Segment the target reply sentence to obtain a target reply segmented sentence;

[0107] Labeling the target approval segmentation sentence according to a preset loan label to obtain a sentence label corresponding to the target approval segmentation sentence;

[0108] Training a preset classification model through the target approval segmentation sentences and the sentence labels;

[0109] The trained classification model is used to perform label classification on the historical reply statements to obtain the reply labels corresponding to the historical reply statements.

[0110] In one embodiment, the reply category statement classification module 102 is specifically used to:

[0111] Matching the reply label with the historical reply sentence to obtain a sentence label mapping relationship;

[0112] Dividing the historical approval statements into pre-loan approval statements, contract approval statements and post-loan approval statements according to the statement label mapping relationship;

[0113] The pre-loan approval statement, the contract approval statement and the post-loan approval statement are determined as the approval category statements.

[0114] In one embodiment, the standard reply opinion generating module 103 is specifically used to:

[0115] Extracting all reply statements in the reply category set;

[0116] Calculate the similarity between adjacent reply statements;

[0117] The reply sentence with the highest similarity is selected as the representative sentence.

[0118] In one embodiment, the standard reply opinion generating module 103 is further used to:

[0119] Extracting the reply template corresponding to the representative sentence;

[0120] Performing semantic analysis on the representative sentences to obtain sentiment semantics and content semantics corresponding to the representative sentences;

[0121] Generate the template content of the reply template according to the emotional semantics and the content semantics;

[0122] The standard reply opinion is generated according to the template content.

[0123] In one embodiment, the approval template filling module 104 is specifically used to:

[0124] Divide the pre-built approval templates into rejection approval templates and approval approval templates according to preset customer needs;

[0125] When the customer demand is an approval demand, the product approval opinions corresponding to the product parameters are selected from the standard approval opinions, and the product approval opinions are added to the approval template;

[0126] When the customer demand is a rejection demand, the customer manager's feedback rejection reason is extracted and the feedback rejection reason is added to the rejection reply template.

[0127] In one embodiment, the standardized reply text generation module 105 is specifically used to:

[0128] Extract the template format of the reply template after adding standard reply opinions;

[0129] Converting the template format into a text format;

[0130] The approval template after adding the standard approval opinions is used to generate a standardized approval text for the target product according to the text format.

[0131] The present invention provides a device for generating standardized reply texts for credit. By extracting some samples from historical approval opinions and marking them, the approval opinions are divided into three module statements, namely, pre-loan, contract, and post-loan, through supervised text classification. The opinions in each module are then subjected to text clustering to obtain the clustered texts. The texts are compared in pairs using a similarity algorithm to obtain representative texts in each class, thereby obtaining pre-loan, contract, and post-loan standard statements. The corresponding reply can be automatically generated with one key according to the historical reply of the account manager and the parameters of the credit product. The reply terms can be accurately filled into the corresponding pre-loan and post-loan modules. The reply opinions are highly standardized, which can effectively improve the communication efficiency between the loan center and the approval center. The loan account manager can perform corresponding operations according to the corresponding terms. The post-loan account manager can strengthen the monitoring of the loan, discover loan anomalies in advance, and reduce the default rate of the loan. Therefore, the method, device, equipment, and medium for generating standardized reply texts for credit proposed by the present invention can solve the problem of low efficiency when performing reply feedback.

[0132] For the specific limitations of the standardized reply text generation device for credit, please refer to the limitations of the intelligent question and answer method above, which will not be repeated here. The various modules in the above-mentioned standardized reply text generation device for credit can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0133] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the service side of a method for generating a standardized approval text for credit.

[0134] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a client side of a method for generating a standardized approval text for credit

[0135] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0136] Obtain historical approval statements of the credit product, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements;

[0137] Performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors;

[0138] Clustering the reply category vectors to obtain a reply category set, selecting representative sentences from the reply category set, and generating standard reply opinions based on the representative sentences;

[0139] Obtain product parameters of the target product, and add the standard approval opinion to a pre-built approval template according to the product parameters;

[0140] Generate a standardized approval text for the target product based on the approval template after adding standard approval opinions.

[0141] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0142] Obtain historical approval statements of the credit product, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements;

[0143] Performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors;

[0144] Clustering the reply category vectors to obtain a reply category set, selecting representative sentences from the reply category set, and generating standard reply opinions based on the representative sentences;

[0145] Obtain product parameters of the target product, and add the standard approval opinion to a pre-built approval template according to the product parameters;

[0146] Generate a standardized approval text for the target product based on the approval template after adding standard approval opinions.

[0147] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0148] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0149] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0150] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit it. If software tools or components other than those of the company appear in the application embodiments, they are only used for illustrative purposes and do not represent actual use. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for generating a standardized approval text for credit granting, characterized in that: include: Obtain historical approval statements of the credit product, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements; Performing text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and converting the reply category sentences into reply category vectors; Clustering the reply category vectors to obtain a reply category set, selecting representative sentences from the reply category set, and generating standard reply opinions based on the representative sentences; Obtain product parameters of the target product, and add the standard approval opinion to a pre-built approval template according to the product parameters; Generate a standardized approval text for the target product based on the approval template after adding standard approval opinions.

2. The method for generating a standardized credit approval text according to claim 1, characterized in that: The labeling of the historical reply statements to obtain the reply labels corresponding to the historical reply statements includes: Selecting a target reply statement from the historical reply statements by a preset random sampling; Segmenting the target reply sentence to obtain a target reply segmented sentence; Labeling the target approval segmentation sentence according to a preset loan label to obtain a sentence label corresponding to the target approval segmentation sentence; Training a preset classification model through the target approval segmentation sentences and the sentence labels; The trained classification model is used to perform label classification on the historical reply statements to obtain the reply labels corresponding to the historical reply statements.

3. The method for generating a standardized credit approval text according to claim 1, characterized in that: The text classification of the historical reply statements according to the reply labels to obtain reply category statements includes: Matching the reply label with the historical reply sentence to obtain a sentence label mapping relationship; Dividing the historical approval statements into pre-loan approval statements, contract approval statements and post-loan approval statements according to the statement label mapping relationship; The pre-loan approval statement, the contract approval statement and the post-loan approval statement are determined as the approval category statements.

4. The method for generating a standardized credit approval text according to claim 1, characterized in that: The selecting of representative sentences in the approval category set includes: Extracting all reply statements in the reply category set; Calculate the similarity between adjacent reply statements; The reply sentence with the highest similarity is selected as the representative sentence.

5. The method for generating a standardized credit approval text according to claim 1, characterized in that: Generating a standard reply opinion according to the representative statement includes: Extracting the reply template corresponding to the representative sentence; Performing semantic analysis on the representative sentences to obtain sentiment semantics and content semantics corresponding to the representative sentences; Generate the template content of the reply template according to the emotional semantics and the content semantics; The standard reply opinion is generated according to the template content.

6. The method for generating a standardized credit approval text according to claim 1, characterized in that: The step of adding the standard approval opinion to a pre-built approval template according to the product parameters includes: Divide the pre-built approval templates into rejection approval templates and approval approval templates according to preset customer needs; When the customer demand is an approval demand, the product approval opinions corresponding to the product parameters are selected from the standard approval opinions, and the product approval opinions are added to the approval template; When the customer demand is a rejection demand, the customer manager's feedback rejection reason is extracted and the feedback rejection reason is added to the rejection reply template.

7. The method for generating a standardized credit approval text according to claim 1, characterized in that: The step of generating a standardized reply text for the target product according to the reply template after adding the standard reply opinion includes: Extract the template format of the reply template after adding standard reply opinions; Converting the template format into a text format; The approval template after adding the standard approval opinions is used to generate a standardized approval text for the target product according to the text format.

8. A device for generating a standardized approval text for credit granting, characterized in that: include: An approval label classification module is used to obtain historical approval statements of credit products, classify the historical approval statements by labels, and obtain approval labels corresponding to the historical approval statements; A reply category sentence classification module is used to perform text classification on the historical reply sentences according to the reply labels to obtain reply category sentences, and convert the reply category sentences into reply category vectors; A standard reply opinion generating module is used to cluster the reply category vectors to obtain a reply category set, select representative sentences from the reply category set, and generate a standard reply opinion based on the representative sentences; An approval template filling module is used to obtain product parameters of a target product and add the standard approval opinions to a pre-built approval template according to the product parameters; The standardized approval text generation module is used to generate the standardized approval text of the target product according to the approval template after adding the standard approval opinions.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for generating a standardized approval text for credit as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for generating a standardized approval text for credit as described in any one of claims 1 to 7 are implemented.

Images