Risk screening method, device, equipment, medium and program product

Abstract

The invention provides a risk screening method which can be applied to the technical field of artificial intelligence or big data. The method comprises the steps of obtaining message data of a customer in response to a transaction request of the customer; performing hit processing on the message data based on a predefined target list to obtain a target message and corresponding hit information, and uploading the target message and the hit information to a pre-trained natural language processing model through a process automation system; judging the hit information by using a natural language processing model to obtain a judgment result of the target message, and transmitting the judgment result back to the process automation system; and marking the transaction processing result of the customer through the process automation system based on the judgment result. The invention further provides a risk screening device and equipment, a storage medium and a program product.

Description

Risk screening methods, devices, equipment, media and procedures Technical Field

[0001] This application relates to the field of artificial intelligence or big data technology, and more specifically to a risk screening method, apparatus, equipment, medium, and program product. Background Technology

[0002] Blacklist screening is a crucial step in ensuring business security during risk screening. However, current blacklist screening systems have certain limitations. Firstly, they still rely on traditional fuzzy matching algorithms. These algorithms lack precision and are prone to misidentifying entities that are not truly on the blacklist, leading to numerous false positives. Secondly, current mainstream message formats do not differentiate message information at the data source level, making it difficult for the screening system to accurately determine the true nature of the entities associated with the messages, further exacerbating the false positive problem. Furthermore, the screening engine has a wide impact on the overall business system; directly modifying the existing system could severely disrupt normal business operations, posing significant and unavoidable risks. Summary of the Invention

[0003] In view of the above problems, this application provides risk screening methods, apparatus, equipment, media and procedures to improve the accuracy and effectiveness of screening.

[0004] According to a first aspect of this application, a risk screening method is provided, the method comprising: in response to a customer's transaction request, acquiring the customer's message data; performing hit processing on the message data based on a predefined target list to obtain target messages and corresponding hit information, and uploading the target messages and the hit information to a pre-trained natural language processing model through a process automation system; using the natural language processing model to judge the hit information to obtain a judgment result of the target messages, and back-transmitting the judgment result to the process automation system; and marking the customer's transaction processing result through the process automation system based on the judgment result.

[0005] According to an embodiment of this application, the method further includes: creating a target interface, the target interface being configured to send the target message and the hit information, and to receive the judgment result of the natural language processing model; and configuring a target process in the process automation system, and calling the target interface through the target process to communicate with the natural language processing model.

[0006] According to an embodiment of this application, the step of using the natural language processing model to judge the hit information and obtain the judgment result of the target message includes: judging whether the type of the hit information is consistent with the type of the target list based on the natural language processing model; and when the type matching is inconsistent, marking the actual type of the hit information and generating the judgment result containing the marking reason.

[0007] According to an embodiment of this application, training the natural language processing model includes: acquiring customer message data within a historical time period and performing data augmentation on the message data to obtain a training sample set; extracting features from the training sample set to obtain target features; inputting the target features into a pre-constructed natural language processing model and obtaining the output of the natural language processing model; calculating the deviation between the output of the natural language processing model and the actual hit rate using a loss function; and when the deviation is greater than a first preset threshold, updating the parameters of the natural language processing model through an optimizer according to the gradient of the loss function until the deviation is less than or equal to the first preset threshold.

[0008] According to an embodiment of this application, the method further includes: obtaining the deviation between the output of the natural language processing model and the actual hit rate; determining that the natural language processing model has converged and stopping the training process when the decreasing trend of the deviation does not reach a second preset threshold within m consecutive training cycles, where m is a positive integer; and using the parameters of the natural language processing model at the time of training termination as the final model parameters.

[0009] According to embodiments of this application, the method further includes: using the process automation system to clean the message data; performing natural language preprocessing on the cleaned message data; and converting the processed message data into a structured feature representation.

[0010] According to an embodiment of this application, marking the customer's transaction processing result through the process automation system based on the judgment result includes: marking the customer's transaction processing result as passed based on the process automation system when the judgment result is that the target message is a false hit; or marking the customer's transaction processing result as rejected based on the process automation system when the judgment result is that the target message is a true hit; and sending the transaction processing result to the reviewer for review.

[0011] A second aspect of this application provides a risk screening device, comprising: an acquisition module, configured to acquire message data of a customer in response to a customer's transaction request; a first screening module, configured to perform hit processing on the message data based on a predefined target list to obtain target messages and corresponding hit information, and to send the target messages and the hit information to a pre-trained natural language processing model through a process automation system; a second screening module, configured to use the natural language processing model to judge the hit information to obtain a judgment result of the target messages, and to transmit the judgment result back to the process automation system; and a determination module, configured to mark the customer's transaction processing result through the process automation system based on the judgment result.

[0012] A third aspect of this application provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the risk screening method described above.

[0013] A fourth aspect of this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the risk screening method described above.

[0014] The fifth aspect of this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the risk screening method described above.

[0015] In the embodiments of this application, a natural language processing model is used to perform secondary filtering on the results screened by the traditional blacklist screening system to determine whether the hit is a false hit, thereby improving the accuracy and effectiveness of risk screening and reducing the processing pressure of false hit warnings. Simultaneously, based on the filtering results, a process automation system is used to process the hit information, directly marking the hit results instead of the original manual marking and submission, thus improving compliance efficiency. Attached Figure Description

[0016] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0017] Figure 1 schematically illustrates an application scenario of the risk screening method, apparatus, device, medium, and program product according to embodiments of this application;

[0018] Figure 2 schematically illustrates a flowchart of training a natural language processing model according to an embodiment of this application;

[0019] Figure 3 schematically illustrates a flowchart of a risk screening method according to an embodiment of this application;

[0020] Figure 4 schematically illustrates a structural block diagram of a risk screening device according to an embodiment of this application; and

[0021] Figure 5 schematically illustrates a block diagram of an electronic device using a risk screening method according to an embodiment of this application. Detailed Implementation

[0022] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0023] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0024] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0025] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0026] In the technical solution of this invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0027] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this application all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0028] Figure 1 schematically illustrates an application scenario of the risk screening method, apparatus, device, medium, and program product according to embodiments of this application.

[0029] As shown in Figure 1, the application scenario 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0030] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0031] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0032] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0033] It should be noted that the risk screening method provided in this application embodiment can generally be executed by server 105. Correspondingly, the risk screening device provided in this application embodiment can generally be located in server 105. The risk screening method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the risk screening device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0034] It should be understood that the number of terminal devices, networks, and servers shown in Figure 1 is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0035] The risk screening method of this application embodiment will be described in detail below based on the scenario described in Figure 1, with reference to Figures 2 and 3.

[0036] Figure 2 schematically illustrates a flowchart of training a natural language processing model according to an embodiment of this application.

[0037] In embodiments of this application, training a natural language processing model includes: acquiring customer message data within a historical time period and performing data augmentation on the message data to obtain a training sample set; extracting features from the training sample set to obtain target features; inputting the target features into a pre-constructed natural language processing model and obtaining the output of the natural language processing model; calculating the deviation between the output of the natural language processing model and the actual hit rate using a loss function; and when the deviation is greater than a first preset threshold, updating the parameters of the natural language processing model through an optimizer according to the gradient of the loss function until the deviation is less than or equal to the first preset threshold.

[0038] When training a natural language processing model, the first step is data preparation, which involves collecting data of different types and sources. For example, remittance message data can be extracted from historical transaction records. This message data can include not only correct blacklist matches but also examples of mismatches to cover various possible mismatch scenarios. Next, the message data can be cleaned to remove invalid or incomplete records. After data cleaning, the data annotation stage can begin. For mismatches, the incorrect parts and the correct matches can be clearly labeled. For correct matches, since the data itself is error-free, it can be used directly without additional annotation.

[0039] Data augmentation techniques, such as synthetic data, rotation, and scaling, can significantly increase the diversity of datasets and improve the generalization ability of models.

[0040] For example, in a bank's blacklist screening system, the core objective is to identify sensitive information about blacklisted customers in transaction messages, such as names, ID numbers, and account numbers. Therefore, new false match examples can be created by synthesizing data to simulate possible false match scenarios. For instance, in the transaction message, the name portion could be changed to content similar to but not identical to a name on the blacklist, and this could be mistakenly identified as a match. Alternatively, modifying the ID number portion of the transaction message to resemble an ID number on the blacklist could also lead to a false match.

[0041] After data augmentation, the dataset can be split into training, validation, and test sets. The training set is used to train the model, the validation set is used to adjust model parameters during training, and the test set is used to evaluate the performance of the trained model.

[0042] Further, the natural language processing (NLP) model training process begins. First, pre-collected and pre-processed message data (such as remittance message datasets) can be loaded, and the text data can be pre-processed, including but not limited to word segmentation, stop word removal, and stemming, to reduce noise and improve model efficiency. Then, features helpful for model learning, such as word frequency, word order, and contextual relationships, can be extracted. If a word embedding-based model architecture is adopted, pre-trained word vectors, such as Word2Vec and GloVe, can be used, or word vectors can be generated from the training data to provide a more targeted semantic representation for the model. Furthermore, a suitable NLP model architecture can be selected based on the data structure and practical application, such as Conditional Random Field (CRF), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and BERT. In this embodiment, considering the characteristics of the message data, Long Short-Term Memory (LSTM) can be selected as the model architecture. After selecting the model, the input and output layers, as well as the intermediate hidden layer structure, can be defined. Next, you can choose the loss function and optimizer. For classification tasks, you can choose the cross-entropy loss function, and simultaneously use the stochastic gradient descent (SGD) algorithm as the optimizer to update model parameters during training. After completing the above settings, you can compile the model, integrating the model architecture, loss function, optimizer, and other information into an executable model, and set the bias, such as accuracy, recall, and F1 score, for subsequent quantitative evaluation of model performance. Then, train the model using the training set data, appropriately setting the batch size and number of training epochs until the bias is less than or equal to a first preset threshold. Simultaneously, you can monitor the model's loss on the training set and its performance on the validation set.

[0043] After the model is trained, its performance can be evaluated on the validation set, and model parameters, such as the learning rate and regularization term, can be adjusted based on the evaluation results.

[0044] Based on the above embodiments, in this embodiment, the method further includes: obtaining the deviation between the output of the natural language processing model and the actual hit rate; determining that the natural language processing model has converged and stopping the training process when the downward trend of the deviation does not reach a second preset threshold within m consecutive training cycles, where m is a positive integer; and using the parameters of the natural language processing model at the time of training termination as the final model parameters.

[0045] For example, an early stopping strategy can be adopted. When the model's performance on the validation set no longer improves or begins to decline, that is, when the decreasing trend of the bias has not reached the second preset threshold within m consecutive training cycles, training can be terminated early, and the parameters of the natural language processing model at the time of training termination can be used as the final model parameters.

[0046] Where m is a positive integer, the specific value of which can be adjusted according to actual training needs and data characteristics. It specifies the number of training rounds allowed without performance improvement, for example, m=8.

[0047] Understandably, terminating training early can prevent natural language processing models from overfitting and improve their generalization ability.

[0048] After evaluation and tuning, the model can be tested on a separate test set to assess its actual performance on unseen data. A thorough analysis of the model's predictions should be conducted to identify potential mismatches. Finally, the trained natural language processing model can be deployed to an accessible server or cloud platform, ensuring it can receive input data and return processing results. Furthermore, a regular feedback loop mechanism can be established to fine-tune the model based on data from manual review by business personnel.

[0049] Understandably, combining new technologies such as natural language processing (NLP) – that is, using NLP models for risk screening – can efficiently and accurately identify risks, quickly process massive amounts of text data, improve screening efficiency and accuracy, and reduce manual costs and false positive rates.

[0050] Figure 3 schematically illustrates a flowchart of a risk screening method according to an embodiment of this application.

[0051] As shown in Figure 3, the risk screening method of this embodiment includes operations S310 to S340.

[0052] In operation S310, in response to the customer's transaction request, the message data of the customer is obtained.

[0053] In embodiments of this application, the customer's consent or authorization can be obtained before acquiring the customer's message data. For example, a request to acquire message data can be sent to the customer before operation S310. If the customer agrees or authorizes the acquisition of message data, operation S310 is performed.

[0054] In operation S320, based on a predefined target list, the message data is processed to obtain the target message and the corresponding hit information. The target message and the hit information are then sent to a pre-trained natural language processing model through a process automation system.

[0055] In operation S330, the natural language processing model is used to judge the hit information, obtain the judgment result of the target message, and transmit the judgment result back to the process automation system.

[0056] In operation S340, based on the judgment result, the customer's transaction processing result is marked by the process automation system.

[0057] In the embodiments of this application, a corresponding operation entry point can be provided to the customer, allowing the customer to choose to agree to or reject the automated decision-making result. That is, before the customer's transaction processing result is marked for processing / decision by the process automation system, the instruction of the customer to agree to or reject the processing / decision can be obtained through the corresponding operation entry point. If the customer agrees to the processing / decision, the customer's transaction processing result is marked for processing / decision by the process automation system, i.e., step S340 is executed. If the customer rejects the processing / decision, the expert decision-making process is initiated.

[0058] In the financial sector, especially in the daily operations of commercial banks, risk screening is a crucial task, directly impacting the safety of bank funds, compliant operations, and the maintenance of customer trust. Currently, the matching algorithms used in mainstream commercial bank blacklist screening systems are outdated. Furthermore, due to limitations in current message submission formats, the wide scope involved, and the significant risks of overall system upgrades, substantial manpower is required to handle false positives, greatly reducing compliance efficiency.

[0059] To address the aforementioned issues, this application provides a risk screening method.

[0060] For example, when a customer initiates a transaction request, the system can respond quickly and obtain the customer's message data (the message receiving and sending processes follow the same screening mechanism, and messages are sent or received according to the normal process). This message data can contain various key transaction information, such as the transaction amount, information of both parties, and the transaction time. Then, the system can perform matching processing on the obtained message data based on a predefined target list. This target list can be pre-defined by the bank according to various risk rules and regulatory requirements, and includes information on entities that may pose a risk, such as blacklisted customers and high-risk areas.

[0061] During the hit processing, the system compares the message data with the target list, filtering out target messages that match the target list and their corresponding hit information. The hit information records which entries in the target list the message data matches, and the specific content of the match. If the blacklist is not matched in the bank's own screening system, this method is not involved, and the message is successfully sent or received. If the blacklist is matched in the bank's own screening system, the target message and hit information can be extracted and sent to a pre-trained natural language processing (NLP) model via a process automation (RPA) system. After receiving the target message and hit information, the NLP model can perform a secondary filtering on the initial target list-based screening. For example, the model can analyze keywords, semantic relationships, and contextual information in the hit information to determine whether the hit truly represents a potential risk, derive the judgment result for the target message, and then send this result back to the process automation system.

[0062] After receiving the judgment results from the natural language processing model, the process automation system can mark the customer's transaction processing results based on the results. If the judgment result is high-risk, the system can mark the transaction as suspicious and trigger corresponding risk warning mechanisms, such as suspending the transaction or requiring the customer to provide additional identity verification information. If the judgment result is low-risk or no-risk, the system can mark the transaction normally and allow it to continue.

[0063] Understandably, utilizing natural language processing models combined with RPA technology, without intruding on the original system, ensures security, reduces the burden of handling false positives in risk screening, improves audit efficiency, and significantly reduces labor costs. Furthermore, it can improve the accuracy of compliance screening and reduce compliance risks in operations.

[0064] In embodiments of this application, the method further includes: using the process automation system to clean the message data; performing natural language preprocessing on the cleaned message data; and converting the processed message data into a structured feature representation.

[0065] By utilizing Process Automation (RPA) systems, acquired message data can be cleaned to remove obviously invalid or missing key information. Simultaneously, to ensure data accuracy and consistency, adhering to a consistent structure and standards, RPA can perform a series of standardization processes, such as removing redundant spaces to avoid identification errors caused by format differences; standardizing naming formats to enable seamless data exchange between different sources; and converting date formats to ensure consistent representation of time information within the system.

[0066] After data cleaning, natural language preprocessing (NLP) can be performed on the message data, including but not limited to word segmentation, part-of-speech tagging, named entity recognition, and stop word removal, to transform the originally complex and disordered text data into a clearly structured format that is easy for the model to process. Finally, the NLP-preprocessed message data can be transformed into a structured feature representation, that is, the rich information in the original text is extracted into a series of features with clear meanings and values, such as word frequency, word vectors, and semantic role labeling results, to jointly constitute the feature vector input to the model.

[0067] Understandably, by introducing RPA for data cleaning and transformation, and by using natural language preprocessing technology to transform text data into structured feature representations, the accuracy and efficiency of risk screening have been significantly improved, providing strong support for compliance screening work.

[0068] In embodiments of this application, the method further includes: creating a target interface, the target interface being configured to send the target message and the hit information, and to receive the judgment result of the natural language processing model; and configuring a target process in the process automation system, and calling the target interface through the target process to communicate with the natural language processing model.

[0069] To achieve efficient automation of the risk screening process and ensure smooth interaction between the automation system and the natural language processing model, a target interface can be created. This interface can be an API (Application Programming Interface) configured to invoke the model and receive its output.

[0070] After creating the target interface, an automated process (i.e., the target process) can be configured in the process automation system. This process can call the deployed target interface to send target messages and hit information to the model and receive the judgment results returned by the model. Specifically, the target process is automatically triggered when the RPA tool reaches a stage that requires interaction with the model. It can call the target interface according to preset logic, send the message data to be processed to the model, and wait for the model to return the processing results. Once the judgment results from the model are received, the target process can immediately feed the results back to the RPA tool so that subsequent operations can be performed based on the results, such as marking the transaction risk level and triggering early warning mechanisms.

[0071] Understandably, by configuring automated processes to call target interfaces, the entire process from data transmission to result reception is automated, thereby not only improving the efficiency of risk screening and reducing the need for manual intervention, but also significantly improving the accuracy and consistency of screening.

[0072] Based on the above embodiments, in this embodiment, the step of using the natural language processing model to judge the hit information and obtain the judgment result of the target message includes: judging whether the type of the hit information is consistent with the type of the target list based on the natural language processing model; and when the type matching is inconsistent, marking the actual type of the hit information and generating the judgment result containing the marking reason.

[0073] After the target message and its hit information are sent to the trained natural language processing model, the model can be used to judge the hit information and determine whether the type of the hit information is consistent with the type defined in the target list. The types in the target list can be predefined based on risk control needs, such as including blacklisted customers and high-risk areas, and each type has its specific attributes and characteristics. When the model finds that the type of the hit information does not match the type of the target list, it can not only determine it as an invalid hit or a false alarm, but also further annotate the relevant hit results according to the judgment results. Specifically, the specific content, context, and potential associations involved in the hit information can be analyzed to accurately label the actual type of the hit information. For example, complex situations may occur, such as address information hitting a name list, name information being incorrectly associated with an address list, or personal identity information being misidentified as an entity or organization; the model can label all of these.

[0074] After completing the actual type labeling, the model can also generate judgment results with detailed labeling reasons. For example, if the matched information is an address but the matched information is a list of names, the labeling reason can be "The address and a certain name in the list have a highly similar expression pattern, which may be due to data entry errors or confusion of related information"; if the matched information is a name but the matched information is an address, the reason can be "The name and a certain address have a specific associated record, but according to the list type definition, this association does not meet the expected type".

[0075] It should be noted that during the processing, the model can continuously interact with the labeled data, adjust its internal parameters through the backpropagation algorithm, optimize model performance, and thus more accurately determine the type of hit information, improve the accuracy of labeling, and thereby enhance the efficiency of the entire risk screening system.

[0076] Understandably, using natural language processing models to accurately identify and annotate hit information can precisely distinguish risk types, reduce misjudgments, and improve the accuracy and efficiency of risk screening.

[0077] Based on the above embodiments, in this embodiment, marking the customer's transaction processing result through the process automation system based on the judgment result includes: marking the customer's transaction processing result as passed based on the process automation system when the judgment result is that the target message is a false hit; or marking the customer's transaction processing result as rejected based on the process automation system when the judgment result is that the target message is a true hit; and sending the transaction processing result to the reviewer for review.

[0078] Once the natural language processing (NLP) model has completed its assessment of the hit information and generated a labeled result, it can transmit the result and related labeling information back to the RPA system via a pre-defined target interface. The RPA then processes the transaction accordingly based on the assessment result. If the NLP model determines that the target message is a false hit (meaning the transaction originally posed no risk but was incorrectly marked as a hit due to data matching issues), the RPA system can automatically mark the customer's transaction processing result as "passed," indicating that it can proceed to the normal processing flow. Conversely, if the NLP model determines that the target message is a true hit (meaning the transaction does indeed have potential risk and matches the risk characteristics in the target list), the RPA system can mark the customer's transaction processing result as "rejected." This mark indicates that the transaction has a high risk and requires further processing; it cannot be directly approved.

[0079] After marking the transaction processing results, the RPA system can send the marked results to reviewers for manual review. For transactions marked "rejected" by the RPA system, reviewers can view the natural language processing model's judgment results, annotation information, and relevant transaction details, and conduct a comprehensive analysis to determine whether the transaction actually poses a risk. If, after review, it is confirmed that the transaction does indeed pose a risk and meets the rejection criteria, the reviewers will confirm the transaction rejection to ensure fund security and prevent potential risks. For transactions marked "false positive" by the RPA system, reviewers can also conduct a comprehensive review. After confirming that the transaction was indeed marked as a false positive, the reviewers can approve the transaction, allowing it to continue normally.

[0080] Understandably, the addition of a manual review process provides a final check on the risk screening results, compensates for the potential limitations of automated systems, and further improves the accuracy and reliability of risk screening.

[0081] Based on the above-described risk screening method, this application also provides a risk screening device. The device will be described in detail below with reference to Figure 4.

[0082] Figure 4 schematically illustrates a structural block diagram of a risk screening device according to an embodiment of this application.

[0083] As shown in Figure 4, the risk screening device 400 of this embodiment includes an acquisition module 410, a first screening module 420, a second screening module 430, and a determination module 440.

[0084] The acquisition module 410 is used to acquire the customer's message data in response to the customer's transaction request. In one embodiment, the acquisition module 410 can be used to perform the operation S310 described above, which will not be repeated here.

[0085] The first screening module 420 is used to perform hit processing on the message data based on a predefined target list, obtain the target message and the corresponding hit information, and send the target message and the hit information to a pre-trained natural language processing model through a process automation system. In one embodiment, the first screening module 420 can be used to perform the operation S320 described above, which will not be repeated here.

[0086] The second screening module 430 is used to use the natural language processing model to judge the hit information, obtain the judgment result of the target message, and transmit the judgment result back to the process automation system. In one embodiment, the second screening module 430 can be used to perform the operation S330 described above, which will not be repeated here.

[0087] The determination module 440 is used to mark the customer's transaction processing result through the process automation system based on the determination result. In one embodiment, the determination module 440 can be used to perform the operation S340 described above, which will not be repeated here.

[0088] According to an embodiment of this application, the first screening module 420 can also be used to: create a target interface, the target interface being configured to send the target message and the hit information, and receive the judgment result of the natural language processing model; and configure a target process in the process automation system, and call the target interface through the target process to communicate with the natural language processing model.

[0089] According to an embodiment of this application, the second screening module 430 is specifically used to: determine whether the type of the hit information is consistent with the type of the target list based on the natural language processing model; and when the type matching is inconsistent, mark the actual type of the hit information and generate the judgment result containing the reason for the markup.

[0090] According to an embodiment of this application, the second screening module 430 can also be used to train the natural language processing model: acquire customer message data within a historical time period, and perform data augmentation on the message data to obtain a training sample set; extract features from the training sample set to obtain target features; input the target features into a pre-built natural language processing model, and obtain the output of the natural language processing model; calculate the deviation between the output of the natural language processing model and the actual hit rate using a loss function; and when the deviation is greater than a first preset threshold, update the parameters of the natural language processing model through an optimizer according to the gradient of the loss function until the deviation is less than or equal to the first preset threshold.

[0091] According to an embodiment of this application, the second screening module 430 can also be used to: obtain the deviation between the output of the natural language processing model and the actual hit rate; when the downward trend of the deviation does not reach a second preset threshold within m consecutive training cycles, determine that the natural language processing model has converged and stop the training process, where m is a positive integer; and use the parameters of the natural language processing model at the time of training termination as the final model parameters.

[0092] According to an embodiment of this application, the acquisition module 410 can also be used to: clean the message data using the process automation system; perform natural language preprocessing on the cleaned message data; and convert the processed message data into a structured feature representation.

[0093] According to an embodiment of this application, the determining module 440 is specifically used to: mark the customer's transaction processing result as passed based on the process automation system when the judgment result is that the target message is a false hit; or mark the customer's transaction processing result as rejected based on the process automation system when the judgment result is that the target message is a true hit; and send the transaction processing result to the reviewer for review.

[0094] According to embodiments of this application, any multiple modules among the acquisition module 410, the first screening module 420, the second screening module 430, and the determination module 440 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the acquisition module 410, the first screening module 420, the second screening module 430, and the determination module 440 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any one of the three implementation methods, or in a suitable combination of any of them. Alternatively, at least one of the acquisition module 410, the first screening module 420, the second screening module 430, and the determination module 440 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0095] Figure 5 schematically illustrates a block diagram of an electronic device suitable for implementing a risk screening method according to an embodiment of this application.

[0096] As shown in FIG. 5, an electronic device 500 according to an embodiment of the present application includes a processor 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage portion 508 into a random access memory (RAM) 503. The processor 501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 501 may also include onboard memory for caching purposes. The processor 501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present application.

[0097] RAM 503 stores various programs and data required for the operation of electronic device 500. Processor 501, ROM 502, and RAM 503 are interconnected via bus 504. Processor 501 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 502 and / or RAM 503. It should be noted that the programs may also be stored in one or more memories other than ROM 502 and RAM 503. Processor 501 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0098] According to embodiments of this application, the electronic device 500 may further include an input / output (I / O) interface 505, which is also connected to a bus 504. The electronic device 500 may also include one or more of the following components connected to the input / output (I / O) interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the input / output (I / O) interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 510 as needed so that computer programs read from it can be installed into the storage section 508 as needed.

[0099] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0100] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503 described above.

[0101] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the risk screening method provided in the embodiments of this application.

[0102] When the computer program is executed by the processor 501, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0103] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 509, and / or installed from a removable medium 511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0104] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by processor 501, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0105] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0107] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

[0108] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of this application, those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.

Claims

1. A risk screening method, characterized in that, The method includes: in response to a customer's transaction request, acquiring the customer's message data; performing hit processing on the message data based on a predefined target list to obtain target messages and corresponding hit information, and sending the target messages and the hit information to a pre-trained natural language processing model through a process automation system; using the natural language processing model to judge the hit information to obtain a judgment result for the target messages, and transmitting the judgment result back to the process automation system; and marking the customer's transaction processing result through the process automation system based on the judgment result.

2. The method according to claim 1, characterized in that, Also includes: Create a target interface, which is configured to send the target message and the hit information, and receive the judgment result of the natural language processing model; The system also configures a target process in the process automation system and calls the target interface through the target process to communicate with the natural language processing model.

3. The method according to claim 1 or 2, characterized in that, The step of using the natural language processing model to judge the hit information and obtain the judgment result of the target message includes: judging whether the type of the hit information is consistent with the type of the target list based on the natural language processing model; and when the type match is inconsistent, marking the actual type of the hit information and generating the judgment result containing the marking reason.

4. The method according to claim 1, characterized in that, Training the natural language processing model includes: acquiring message data from customers within a historical time period and performing data augmentation on the message data to obtain a training sample set; extracting features from the training sample set to obtain target features; inputting the target features into a pre-constructed natural language processing model and obtaining the output of the natural language processing model; calculating the deviation between the output of the natural language processing model and the actual hit rate using a loss function; and when the deviation is greater than a first preset threshold, updating the parameters of the natural language processing model through an optimizer according to the gradient of the loss function until the deviation is less than or equal to the first preset threshold.

5. The method according to claim 4, characterized in that, Also includes: Obtain the deviation between the output of the natural language processing model and the actual hit rate; If the decreasing trend of the deviation does not reach the second preset threshold within m consecutive training cycles, the natural language processing model is determined to have converged and the training process is stopped, where m is a positive integer; And the parameters of the natural language processing model at the end of training are used as the final model parameters.

6. The method according to claim 1 or 4, characterized in that, It also includes: using the process automation system to clean the message data; and performing natural language preprocessing on the cleaned message data, and converting the processed message data into a structured feature representation.

7. The method according to claim 1, characterized in that, The step of marking the customer's transaction processing result through the process automation system based on the judgment result includes: marking the customer's transaction processing result as passed based on the process automation system when the judgment result is that the target message is a false hit; or marking the customer's transaction processing result as rejected based on the process automation system when the judgment result is that the target message is a true hit; and sending the transaction processing result to the reviewer for review.

8. A risk screening device, characterized in that, The device includes: an acquisition module, configured to acquire the customer's message data in response to a customer's transaction request; a first screening module, configured to perform hit processing on the message data based on a predefined target list to obtain target messages and corresponding hit information, and to send the target messages and the hit information to a pre-trained natural language processing model through a process automation system; a second screening module, configured to use the natural language processing model to judge the hit information to obtain a judgment result of the target messages, and to transmit the judgment result back to the process automation system; and a determination module, configured to mark the customer's transaction processing result through the process automation system based on the judgment result.

9. An electronic device, comprising: One or more processors; A storage device for storing one or more computer programs, characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images