Suspicious transaction report generation method, device and equipment
By combining the analysis of suspicious transaction report template structure with a hallucination detection model, suspicious transaction reports are automatically generated, solving the problem of low efficiency in traditional methods and achieving efficient and accurate report generation with strong adaptability.
Patent Information
- Application Number
- CN202511771413.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-28
- Publication Date
- 2026-04-07
AI Technical Summary
Traditional suspicious transaction report generation processes rely on expert rule identification and manual analysis, which is inefficient, difficult to meet the needs of real-time data processing, and prone to human error.
By analyzing the structure of suspicious transaction report templates, using automated tools and expert rules to screen suspicious transaction clues, and combining them with an illusion detection model to generate report content, the accuracy and completeness of the reports are ensured.
It enables the automated generation of suspicious transaction reports, improving generation efficiency, reducing manual intervention, ensuring the accuracy and compliance of reports, and is highly adaptable to cope with new risks and regulatory changes.
Smart Images
Figure CN121809434A_ABST
Abstract
Description
Technical Field
[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to a method, apparatus, and device for generating suspicious transaction reports. Background Technology
[0002] In modern financial institutions, the analysis of suspicious transactions is facing challenges due to the massive scale and increasing complexity of transaction data, as well as increasingly stringent compliance requirements. Traditional Suspicious Transaction Report (STR) generation processes primarily rely on suspicious transaction leads identified based on expert rules, followed by manual analysis and report writing. This process is inefficient, struggles to handle real-time data processing demands, and is prone to human error. Summary of the Invention
[0003] To address the problems existing in the prior art, embodiments of this specification provide a method, apparatus, and device for generating suspicious transaction reports. The method analyzes the structure of a suspicious transaction report template to obtain multiple fill-in items and the generation order of the content within each item. Following this generation order, it performs in-depth analysis of suspicious transaction clues selected by automated tools for generating the content of each fill-in item and expert rules, obtaining the content of each fill-in item. The content of each fill-in item is then verified for factuality using an illusion detection model. The content of each fill-in item is then entered into the suspicious transaction report template to obtain the suspicious transaction report. The method of this specification achieves automated generation of suspicious transaction reports, replacing the traditional method of manually writing suspicious transaction reports, avoiding human error, and improving the efficiency of suspicious transaction report generation.
[0004] The specific technical solutions of the embodiments in this specification are as follows:
[0005] On the one hand, embodiments of this specification provide a method for generating a suspicious transaction report, the method comprising:
[0006] The structure of the suspicious transaction report template was analyzed to obtain multiple fill-in items and the generation order of the content in each fill-in item;
[0007] According to the generation order, each fill item is analyzed in depth based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering transaction data in advance through expert rules, so as to obtain the content of each fill item.
[0008] The content of each fill-in item is factually verified using an illusion detection model;
[0009] After the factual checks of all fields have passed, the contents of each field are entered into the suspicious transaction report template to obtain the suspicious transaction report.
[0010] Furthermore, analysis of the structure of the suspicious transaction report template reveals multiple fill-in items and the generation order of the content within each item, including:
[0011] A first prompt word template is generated based on the content of the suspicious transaction report template and the position of each of the fill-in items in the suspicious transaction report template;
[0012] The first prompt word template is input into a pre-trained large model for calculation to obtain the first planning content, which includes the first prompt word of each fill item and the generation order of each fill item.
[0013] Furthermore, following the aforementioned generation order, each fill item is subjected to in-depth analysis based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by pre-screening transaction data through expert rules. The content of each fill item further includes:
[0014] Following the generation order, a second prompt word template for each filler item is generated sequentially based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues.
[0015] The second prompt word template is input into the large model for calculation to obtain the content of each fill item.
[0016] Furthermore, if the first prompt word of a later filler item in the generation order includes at least one of the previous filler items, generating a second prompt word template for each filler item according to the generation order, based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues, further includes:
[0017] For the next filler item in the generation sequence, a second prompt word template for the filler item is generated based on the first prompt word of the filler item, the corresponding automated tool, the suspicious transaction clue, and the content of at least one previous filler item.
[0018] Furthermore, if the factual verification of the content of the fill item fails, the method further includes:
[0019] The hallucination detection model outputs the reason why the factual detection failed;
[0020] Based on the first prompt word of the fill-in item that failed the factual detection, the corresponding automated tool, the suspicious transaction clue, and the reason for the factual detection failure, a second prompt word template for each fill-in item is regenerated.
[0021] The regenerated second prompt word template is input into the large model for calculation, and the content of the fill item is obtained again.
[0022] Furthermore, the fields to be filled in include actual customer information, customer information analysis results, and brief descriptions;
[0023] The generation sequence is as follows: first, generate the content of the customer's actual information; then, generate the customer information analysis result based on the content of the customer's actual information; and finally, generate the summary information based on the customer's actual information and the customer information analysis result.
[0024] On the other hand, embodiments of this specification also provide an apparatus for generating suspicious transaction reports, the apparatus comprising:
[0025] The template analysis unit is used to analyze the structure of the suspicious transaction report template to obtain multiple fill items and the generation order of the content in each fill item;
[0026] The fill item analysis unit is used to perform in-depth analysis on each fill item according to the generation order, based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering the transaction data in advance through expert rules, so as to obtain the content of each fill item.
[0027] The hallucination detection unit is used to perform factual detection on the content of each filler item using a hallucination detection model;
[0028] The report filling unit is used to fill in the contents of each fill-in item into the suspicious transaction report template after the factual checks of the contents of each fill-in item have passed, so as to obtain a suspicious transaction report.
[0029] On the other hand, embodiments of this specification also provide a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the above-described method.
[0030] On the other hand, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.
[0031] On the other hand, embodiments of this specification also provide a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method.
[0032] The technical solutions in the embodiments of this specification have achieved significant effects in several aspects:
[0033] 1. Improved the level of automation in report generation: Through automated analysis, manual intervention has been greatly reduced, enabling the automatic generation of suspicious transaction reports and significantly improving work efficiency.
[0034] 2. Enhanced data processing and integration capabilities: The methods in the embodiments of this specification can efficiently integrate multiple data sources, including customer information, transaction records, and external databases, to ensure the completeness and accuracy of the report content.
[0035] 3. Improved accuracy and reliability of reports: By introducing hallucination detection technology, the generated report content is ensured to be based on real data, effectively avoiding the occurrence of false information or false reports.
[0036] 4. Enhanced adaptability and scalability: The system can respond promptly to new risks and ever-changing regulatory requirements, and supports flexible expansion, making it suitable for various complex scenarios and ensuring continuous improvement in compliance and risk management capabilities.
[0037] 5. Optimized workflow: The methods in the embodiments of this specification greatly simplify the report generation process, reduce the burden on compliance and audit personnel, and enable them to focus on higher-level analysis and decision-making. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 The diagram shown is a schematic representation of an implementation system for a method of generating a suspicious transaction report according to an embodiment of this specification.
[0040] Figure 2 The diagram shown is a flowchart illustrating the analysis of the structure of the suspicious transaction report template in the embodiments of this specification, resulting in multiple fill-in items and the generation order of the content in each fill-in item;
[0041] Figure 3 The diagram shown is a flowchart illustrating the process in this embodiment of the specification, in which each filler item is analyzed in depth according to the generation order described above, based on the automated tools corresponding to each filler item and the suspicious transaction clues obtained by filtering transaction data in advance through expert rules, to obtain the content of each filler item.
[0042] Figure 4 The diagram shown is a flowchart illustrating the process of regenerating the content of a fill-in item when the factual detection of the fill-in item fails in an embodiment of this specification.
[0043] Figure 5 The diagram shown is a structural schematic of a suspicious transaction report generation device according to an embodiment of this specification.
[0044] Figure 6 The diagram shown is a structural schematic of the computer device in an embodiment of this specification.
[0045] [Explanation of Figure Markers]:
[0046] 501. Template Analysis Unit;
[0047] 502. Fill-in Item Analysis Unit;
[0048] 503. Hallucination Detection Unit;
[0049] 504. Report Filling Unit;
[0050] 602. Computer equipment;
[0051] 604, Processor;
[0052] 606. Memory;
[0053] 608. Drive mechanism;
[0054] 610. Input / output module;
[0055] 612. Input devices;
[0056] 614. Output devices;
[0057] 616. Presentation equipment;
[0058] 618. Graphical User Interface;
[0059] 620. Network interface;
[0060] 622. Communication link;
[0061] 624. Communication bus. Detailed Implementation
[0062] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the embodiments of this specification.
[0063] It should be noted that the terms "first," "second," etc., in the description, claims, and accompanying drawings of the embodiments herein are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0064] It should be noted that the acquisition, storage, use, and processing of data in the technical solutions of the embodiments of this specification all comply with the relevant provisions of national laws and regulations.
[0065] It should be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.
[0066] To address the inefficiency of traditional Suspicious Transaction Report (STR) generation processes, which rely heavily on expert-rule-based identification of suspicious transaction leads followed by manual analysis and report writing, this specification provides a method for generating suspicious transaction reports. This method is inefficient, struggles to handle real-time data processing demands, and is prone to human error. Figure 1 The diagram illustrates a flowchart of a method for generating a suspicious transaction report according to an embodiment of this specification. While the process of generating a suspicious transaction report is described in this diagram, it can include more or fewer steps based on conventional or non-creative labor. The order of steps listed in the embodiment is merely one possible execution order among many and does not represent the only possible order. In actual system or device products, the methods shown in the embodiment or the accompanying drawings can be executed sequentially or in parallel. Specifically, as shown... Figure 1 As shown, the method can be executed by a server and may include:
[0067] Step 101: Analyze the structure of the suspicious transaction report template to obtain multiple fill-in items and the generation order of the content in each fill-in item;
[0068] Step 102: According to the generation order, perform in-depth analysis on each fill item based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering transaction data in advance through expert rules, and obtain the content of each fill item;
[0069] Step 103: Perform factual checks on the content of each fill-in item using an illusion detection model;
[0070] Step 104: After the factual checks of the contents of each fill-in item have passed, fill in the contents of each fill-in item in the suspicious transaction report template to obtain the suspicious transaction report.
[0071] In this embodiment, the suspicious transaction report template can be predefined by staff, including the format and structure of the suspicious transaction report. Multiple preset slots are set in the suspicious transaction report template, each representing a fill item. The content to be filled in the preset slots is the content of the corresponding fill item analyzed later. By using the positions of the preset slots, the subsequently analyzed content is embedded into the fixed suspicious transaction report template, thereby ensuring that the generated suspicious transaction report conforms to industry standards and requirements in terms of structure and presentation. Each slot in the suspicious transaction report template corresponds to a different type of information, ensuring consistency and readability of the report output.
[0072] For example, the preset slots in the suspicious transaction report template can be represented by “[]” or other characters. If the preset slot is represented by the character “[]”, then the content to be filled in “[]” is the content of the corresponding fill item.
[0073] In some other embodiments of this specification, the names of fill-in items can be pre-written within the brackets "[]" in the suspicious transaction report template to facilitate subsequent determination of the fill-in items and analysis of their content. For example, a portion of the content in the suspicious transaction report template is as follows:
[0074] "Today, our bank discovered [a brief summary of suspicious transactions]. The relevant information is reported as follows:"
[0075] I. Basic Customer Information
[0076] [Basic information, including personal details, account opening documents, characteristics, and activity history.]
[0077] II. Fund Transaction Details
[0078] (I) General Situation
[0079] [Time, Number of Transaction Details, Amount, etc.]
[0080] (II) Sources of Funds
[0081] [Funding Source Information]
[0082] (III) Where the funds went
[0083] [Where Funds Have Been Used]
[0084] III. Analysis of Suspicious Points
[0085] [Trading Characteristics, Key Accounts]
[0086] IV. Overall Conclusion
[0087] Based on data analysis and past experience, identify the types of violations that may be involved.
[0088] This is to report.
[0089] Time: [Report Generation Time]
[0090] The suspicious transaction report template in the above example includes the following fields: "Brief Overview of Suspicious Transactions," "Basic Information, including account holder's personal information, account opening materials, characteristics, and activity details," "Time, Number of Transactions, Amount," "Fund Source Information," "Fund Destination Information," "Transaction Characteristics and Key Accounts," "Based on Data Analysis and Past Experience, a Judgment of the Type of Regulations That May Be Violated," and "Report Generation Time." This embodiment requires first analyzing the generation order of the content in these fields.
[0091] Specifically, such as Figure 2 As shown, the structure of the suspicious transaction report template is analyzed, revealing multiple fill-in items and the generation order of the content within each item, which further includes:
[0092] Step 201: Generate a first prompt word template based on the content of the suspicious transaction report template and the position of each of the fill-in items in the suspicious transaction report template;
[0093] Step 202: Input the first prompt word template into the pre-trained large model for calculation to obtain the first planning content, which includes the first prompt word of each fill item and the generation order of each fill item.
[0094] In the embodiments of this specification, the generation order of fill-in items can be analyzed using a large model. The large model can be a general large model. First, a first prompt word template needs to be generated, which includes the position of each fill-in item in the suspicious transaction report template. For example, the first prompt word template is: You are a bank specialist responsible for reporting suspicious transactions. Please read the suspicious transaction report template. The content to be generated is in "[]". Please formulate a generation plan.
[0095] The generated first prompt word template is then input into the large model for calculation. The large model will output the first planning content, which includes the first prompt word for each fill item and the generation order of each fill item.
[0096] For example, the content of the first plan could be:
[0097] Step 1: Generate basic customer information, including account holder information, account opening documents, characteristics, and activity details;
[0098] Step 2: Generate transaction details, including time, number of transactions, amount, source of funds, and destination of funds;
[0099] Step 3: Generate suspicious activity analysis, including transaction characteristics, whether it is a key account, etc.;
[0100] Step 4: Generate a comprehensive conclusion, combining data analysis and past experience, and determine the type of violation that may be involved;
[0101] Step 5: Generate a concise overview of the suspicious transaction, summarizing the generated content to form a brief overview of the entire suspicious transaction event;
[0102] Step 6: Extract the current time.
[0103] Then, following the generation order, each filler item is analyzed in depth using the automated tools corresponding to each filler item and the suspicious transaction clues obtained by filtering transaction data in advance through expert rules, so as to obtain the content of each filler item.
[0104] In the embodiments described in this specification, such as Figure 3 As shown, following the generation order, each fill item is analyzed in depth using the automated tools corresponding to each fill item and the suspicious transaction clues obtained by pre-screening transaction data through expert rules. The content of each fill item further includes:
[0105] Step 301: Following the generation order, generate a second prompt word template for each filler item based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues;
[0106] Step 302: Input the second prompt word template into the large model for calculation to obtain the content of each fill item.
[0107] In the embodiments of this specification, the content of each fill item can be analyzed using a large model.
[0108] The fill-in items in this embodiment include three categories: actual customer information, customer information analysis results, and brief information. Each category may include one or more specific items. Since, for example, the fill-in item for actual customer information requires obtaining the customer's actual information, and obtaining this information requires appropriate tools, this embodiment pre-configures a toolbox. The toolbox integrates various tools, supporting query functions for customer information, transaction records, and the knowledge base, and also has a code executor function. Querying basic customer information allows obtaining detailed data related to the customer; querying transaction records allows the system to access and analyze specific transaction details; querying the knowledge base provides background information and industry knowledge support related to suspicious transactions; and the code executor can run custom scripts or code when needed, further expanding the system's analytical capabilities.
[0109] Therefore, according to the embodiments in this specification Figure 2 The method described includes the first prompt word for the fill-in item, the corresponding automated tool, and the second prompt word template generated from pre-analyzed suspicious transaction clues. The suspicious transaction clues are based on expert rules used to filter suspicious activities in transaction data. Through a predefined set of rules, this module can identify potential suspicious transaction behaviors from a large number of transactions and use these clues as input for subsequent processing by the system. This embodiment also incorporates suspicious transaction clues into the prompt word template for generating fill-in item content, guiding large models to perform precise analysis and improving accuracy.
[0110] For example, for the fill-in-the-blank field "Customer Basic Information", the second prompt template could be:
[0111] Please customize the plan to generate basic customer information. This will give you access to the following tools: {a collection of automated tools corresponding to the "Basic Customer Information" field} and suspicious transaction leads {suspicious transaction leads filtered based on expert rules}.
[0112] After inputting the second cue word template into the large model, the large model will generate planning content (i.e., the large model's thought process). For example, the planning content generated for the second cue word template corresponding to the fill-in item "Customer Basic Information" mentioned above could be:
[0113] To generate basic customer information, you need to use a basic customer information query tool to obtain personal information and then organize the text paragraphs.
[0114] In the embodiments of this specification, the order in which the content of the fill-in items is generated for the three categories of actual customer information, customer information analysis results, and brief information is as follows: first, the content of actual customer information is generated; then, the customer information analysis results are generated based on the content of actual customer information; and finally, the brief information is generated based on the actual customer information and the customer information analysis results. Therefore, when generating the content of the fill-in item "Customer Information Analysis Results," the content of the fill-in item "Basic Customer Information" is required as input; when generating the content of the fill-in item "Brief Information," the content of both the basic customer information and the customer information analysis results are required as input.
[0115] In the embodiments of this specification, the above-mentioned input relationships are all reflected in the results of the large model's analysis of the first prompt word template. That is, the first prompt word of the later filler item in the generation order may include at least one previous filler item (if it includes two or more previous filler items, it includes two or more specified filler items ordered before the current filler item). In this case, according to the generation order, generating the second prompt word template of each filler item in sequence based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues further includes:
[0116] For the next filler item in the generation sequence, a second prompt word template for the filler item is generated based on the first prompt word of the filler item, the corresponding automated tool, the suspicious transaction clue, and the content of at least one previous filler item.
[0117] For example, the first prompt for this fill-in-the-blank item includes: a summary of the customer's basic information, fund transaction details, analysis of suspicious points, and a comprehensive conclusion, forming a brief overview of the entire suspicious transaction event.
[0118] The second cue word template for this fill-in-the-blank item can be briefly summarized as follows:
[0119] Please gather basic customer information, transaction details, analysis of suspicious points, and comprehensive conclusions to generate a brief overview of the suspicious transaction event. This will provide you with the following tools: {a collection of automated tools corresponding to the "brief overview" field} and suspicious transaction leads {suspicious transaction leads filtered based on expert rules}.
[0120] In the embodiments of this specification, because the large model may generate non-existent facts or false statements based on incomplete or misunderstanding information when generating content, the embodiments of this specification, when generating the content of each fill item in sequence, will also perform factual detection on the output results of the large model through a pre-trained illusion detection model after the large model outputs the results, so as to prevent the large model from generating non-existent facts or false statements.
[0121] Optionally, the hallucination detection model can be LLM-as-a-Judge, HHEM, etc., and the embodiments in this specification are not limited thereto.
[0122] Specifically, such as Figure 4 As shown, if the factual verification of the content of the fill item fails, the method further includes:
[0123] Step 401: The hallucination detection model outputs the reason why the factual detection failed;
[0124] Step 402: Regenerate the second prompt word template for each fill-in item based on the first prompt word of the fill-in item that failed the factual detection, the corresponding automated tool, the suspicious transaction clue, and the reason for the factual detection failure;
[0125] Step 403: Input the regenerated second prompt word template into the large model for calculation to obtain the content of the fill item again.
[0126] In this embodiment of the specification, if the hallucination detection model fails the factual detection of the content generated by the large model, it will output the reason for the failure. This embodiment of the specification guides the large model to regenerate content based on the reason for the factual detection failure until the factual detection result of the content generated by the large model passes, and then uses that content as the final content of the current fill item. Then, the analysis of the next fill item is performed in sequence.
[0127] After the factual checks of all fields have passed, the contents of each field are entered into the suspicious transaction report template to obtain the suspicious transaction report.
[0128] In some other embodiments of this specification, after obtaining the content of the previous fill item, key information such as important data and core conclusions can be extracted from the content of the previous fill item. After the content of the next fill item (whose first prompt includes the previous fill item) is generated, the key information of the next fill item is extracted, the consistency between the key information of the previous fill item and the key information of the next fill item is calculated, and if there is still a discrepancy, the content of the next fill item is regenerated.
[0129] This can be understood as follows: by using the above method, in cases where the report is very long, it can avoid inconsistencies caused by the large model forgetting the content of the previous filler item when generating the content of the next filler item.
[0130] Based on the same inventive concept, embodiments of this specification also provide an apparatus for generating suspicious transaction reports, such as... Figure 5 As shown, the device includes:
[0131] The template analysis unit 501 is used to analyze the structure of the suspicious transaction report template to obtain multiple fill items and the generation order of the content in each fill item;
[0132] The fill item analysis unit 502 is used to perform in-depth analysis on each fill item according to the generation order, based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering the transaction data in advance through expert rules, so as to obtain the content of each fill item.
[0133] The hallucination detection unit 503 is used to perform factual detection on the content of each fill-in item through the hallucination detection model;
[0134] The report filling unit 504 is used to fill in the contents of each fill item in the suspicious transaction report template after the factual checks of the contents of each fill item have passed, so as to obtain a suspicious transaction report.
[0135] The beneficial effects obtained by the above-described device are the same as those obtained by the above-described method, and will not be described in detail in the embodiments of this specification.
[0136] like Figure 6 The diagram shown is a structural schematic of a computer device according to an embodiment of this specification. The methods described in this specification can be applied to the computer device of this embodiment.
[0137] Computer device 602 may include one or more processors 604, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. Computer device 602 may also include any memory 606 for storing information of any kind, such as code, settings, data, etc. Non-limitingly, for example, memory 606 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any storage resource can be used to store information using any technology.
[0138] Furthermore, any storage resource can provide volatile or non-volatile retention of information.
[0139] Furthermore, any storage resource can represent a fixed or removable component of the computer device 602. In one case, when the processor 604 executes associated instructions stored in any storage resource or combination of storage resources, the computer device 602 can perform any operation of the associated instructions. The computer device 602 also includes one or more drive mechanisms 608 for interacting with any storage resource, such as a hard disk drive system, an optical disk drive system, etc.
[0140] Computer device 602 may also include an input / output module 610 (I / O) for receiving various inputs (via input device 612) and providing various outputs (via output device 614). A specific output mechanism may include a presentation device 616 and an associated graphical user interface (GUI) 618. In other embodiments, the input / output module 610 (I / O), input device 612, and output device 614 may be omitted, and the device may function solely as a computer device within a network. Computer device 602 may also include one or more network interfaces 620 for exchanging data with other devices via one or more communication links 622. One or more communication buses 624 couple the components described above together.
[0141] Communication link 622 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 622 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.
[0142] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.
[0143] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the above-described method.
[0144] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply 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 embodiments of this specification.
[0145] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the embodiments of this specification, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0146] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the embodiments in this specification.
[0147] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0148] In the embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.
[0149] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.
[0150] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0151] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this specification, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0152] This specification describes the principles and implementation methods of the embodiments using specific examples. The above descriptions of the embodiments are only for the purpose of helping to understand the methods and core ideas of the embodiments in this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the embodiments in this specification. Therefore, the content of this specification should not be construed as a limitation on the embodiments in this specification.
Claims
1. A method for generating a suspicious transaction report, characterized in that, The method includes: The structure of the suspicious transaction report template was analyzed to obtain multiple fill-in items and the generation order of the content in each fill-in item; According to the generation order, each fill item is analyzed in depth based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering transaction data in advance through expert rules, so as to obtain the content of each fill item. The content of each fill-in item is factually verified using an illusion detection model; After the factual checks of all fields have passed, the contents of each field are entered into the suspicious transaction report template to obtain the suspicious transaction report.
2. The method according to claim 1, characterized in that, Analyzing the structure of the suspicious transaction report template reveals multiple fill-in items and the generation order of the content within each item, which further includes: A first prompt word template is generated based on the content of the suspicious transaction report template and the position of each of the fill-in items in the suspicious transaction report template; The first prompt word template is input into a pre-trained large model for calculation to obtain the first planning content, which includes the first prompt word of each fill item and the generation order of each fill item.
3. The method according to claim 2, characterized in that, Following the aforementioned generation order, each fill item is subjected to in-depth analysis using the automated tools corresponding to each fill item and suspicious transaction clues obtained by pre-screening transaction data through expert rules. The content of each fill item further includes: Following the generation order, a second prompt word template for each filler item is generated sequentially based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues. The second prompt word template is input into the large model for calculation to obtain the content of each fill item.
4. The method according to claim 3, characterized in that, If the first prompt word of a later filler item in the generation order includes at least one of the previous filler items, generating a second prompt word template for each filler item according to the generation order, based on the first prompt word of each filler item, the corresponding automated tool, and the suspicious transaction clues, further includes: For the next filler item in the generation sequence, a second prompt word template for the filler item is generated based on the first prompt word of the filler item, the corresponding automated tool, the suspicious transaction clue, and the content of at least one previous filler item.
5. The method according to claim 3, characterized in that, If the factual verification of the content of the fill item fails, the method further includes: The hallucination detection model outputs the reason why the factual detection failed; Based on the first prompt word of the fill-in item that failed the factual detection, the corresponding automated tool, the suspicious transaction clue, and the reason for the factual detection failure, a second prompt word template for each fill-in item is regenerated. The regenerated second prompt word template is input into the large model for calculation, and the content of the fill item is obtained again.
6. The method according to claim 3, characterized in that, The fields to be filled in include actual customer information, customer information analysis results, and brief information; The generation sequence is as follows: first, generate the content of the customer's actual information; then, generate the customer information analysis result based on the content of the customer's actual information; and finally, generate the summary information based on the customer's actual information and the customer information analysis result.
7. An apparatus for generating suspicious transaction reports, characterized in that, The device includes: The template analysis unit is used to analyze the structure of the suspicious transaction report template to obtain multiple fill items and the generation order of the content in each fill item; The fill item analysis unit is used to perform in-depth analysis on each fill item according to the generation order, based on the automated tools corresponding to each fill item and the suspicious transaction clues obtained by filtering the transaction data in advance through expert rules, so as to obtain the content of each fill item. The hallucination detection unit is used to perform factual detection on the content of each filler item using a hallucination detection model; The report filling unit is used to fill in the contents of each fill-in item into the suspicious transaction report template after the factual checks of the contents of each fill-in item have passed, so as to obtain a suspicious transaction report.
8. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.
10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.