Transaction verification method, device, equipment, medium and program product

Abstract

The invention provides a transaction verification method which can be applied to the technical field of artificial intelligence and the technical field of financial science and technology. The transaction verification method comprises the steps that in response to acquisition of a transaction verification book, the type of the transaction verification book is determined according to transaction information of the transaction verification book, the transaction verification book is sent by a transaction object, and the transaction information comprises at least one of the transaction object and a transaction type; determining a target cue word from a preset cue word set based on the type of the transaction confirmation book; obtaining a target transaction element in the transaction confirmation book based on the target cue word by using a large model; and matching the target transaction element with a preset candidate transaction element, and performing transaction verification on the transaction verification book according to a matching result. The invention also provides a transaction verification device, equipment, a medium and a program product.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and the technical field of financial technology, and more particularly to a transaction verification method, device, equipment, medium and program product. BACKGROUND

[0002] In the current transaction processing process, the transaction object sends a transaction verification letter by email, and the transaction verification letter in the email is usually downloaded by the background operation personnel and imported into the system. The transaction elements in the transaction verification letter are extracted by natural language processing technology and compared with the internal transaction record to determine whether the content in the transaction verification letter sent by the transaction object is consistent with the internal transaction record. However, the verification letter templates used by different transaction counterparts differ greatly in format, field naming and expression habits, resulting in low accuracy of transaction element extraction and automatic matching, and a large number of abnormalities need to be manually reviewed, which affects the transaction confirmation efficiency and increases the operational risk. SUMMARY

[0003] In view of the above problems, the present application provides a transaction verification method, device, equipment, medium and program product.

[0004] According to a first aspect of the present application, a transaction verification method is provided, comprising: in response to obtaining a transaction verification letter, determining the category of the transaction verification letter according to the transaction information of the transaction verification letter, the transaction verification letter being sent by a transaction object, the transaction information including at least one of a transaction object and a transaction type; determining a target prompt word from a preset prompt word set based on the category of the transaction verification letter; obtaining a target transaction element in the transaction verification letter based on the target prompt word using a large model; matching the target transaction element with a preset candidate transaction element, and verifying the transaction verification letter based on the matching result.

[0005] According to an embodiment of the present application, in response to obtaining a transaction verification letter, determining the category of the transaction verification letter according to the transaction information of the transaction verification letter includes: in response to obtaining the transaction verification letter from a target mailbox, determining the attribute information of the target mailbox, the attribute information including at least one of the sender domain name, the mailbox subject, the attachment type and the format features of the transaction verification letter; extracting at least one of the transaction object and the transaction type from the attribute information as the transaction information of the transaction verification letter; processing the transaction information to obtain the category of the transaction verification letter.

[0006] According to an embodiment of the present application, the transaction information is processed to obtain the category of the transaction certificate, including: obtaining a plurality of category information in a preset category set, the category information including at least one of a transaction object and a transaction type; determining the category information matched with the transaction information from the plurality of category information as the category of the transaction certificate; or calculating the semantic similarity between the transaction information and each of the plurality of category information, and determining the category information with a semantic similarity higher than a preset similarity threshold in the plurality of category information as the category of the transaction certificate.

[0007] According to an embodiment of the present application, the prompt word set stores a plurality of candidate categories and candidate prompt words corresponding to each candidate category; and the target prompt word is determined from the preset prompt word set based on the category of the transaction certificate, including: determining the candidate category matched with the category of the transaction certificate from the plurality of candidate categories as the target category; and determining the candidate prompt word corresponding to the target category in the prompt word set as the target prompt word.

[0008] According to an embodiment of the present application, the target transaction element in the transaction certificate is obtained based on the target prompt word using a large model, including: performing layout analysis and character recognition on the transaction certificate to obtain a multi-modal feature of the transaction certificate, the multi-modal feature including text content and layout features, the layout features including at least one of position information of each text content in the transaction certificate, table structure, and visual layout features; and obtaining the target transaction element in the transaction certificate based on the target prompt word and the multi-modal feature using the large model.

[0009] According to an embodiment of the present application, the target transaction element in the transaction certificate is obtained based on the target prompt word using a large model, including: obtaining a plurality of to-be-evaluated transaction elements in the transaction certificate based on the target prompt word using the large model; performing verification on each to-be-evaluated transaction element based on a plurality of preset verification strategies, and determining a confidence of each to-be-evaluated transaction element based on a verification result; and processing the plurality of to-be-evaluated transaction elements based on the confidence to obtain the target transaction element.

[0010] According to an embodiment of the present application, the target transaction element is obtained by processing the plurality of to-be-evaluated transaction elements based on the confidence, including: for any to-be-evaluated transaction element in the plurality of to-be-evaluated transaction elements, performing the following operations: in response to the confidence of the to-be-evaluated transaction element being greater than or equal to a first threshold, taking the to-be-evaluated transaction element as the target transaction element; in response to the confidence of the to-be-evaluated transaction element being less than the first threshold and greater than a second threshold, taking the to-be-evaluated transaction element as the target transaction element in a case where confirmation information for the to-be-evaluated transaction element is obtained; wherein the second threshold is less than the first threshold; in response to the confidence of the to-be-evaluated transaction element being less than or equal to the second threshold, evaluating the to-be-evaluated transaction element based on the target prompt word and the transaction certificate, and taking the to-be-evaluated transaction element as the target transaction element in a case where an evaluation result meets a predetermined condition.

[0011] According to an embodiment of the present application, before the to-be-evaluated transaction element is evaluated based on the target prompt word and the transaction certificate, the position information of the to-be-evaluated transaction element in the transaction certificate is determined, and the target prompt word is expanded based on the position information.

[0012] According to an embodiment of the present application, the evaluation of the to-be-evaluated transaction element based on the target prompt word and the transaction certificate includes: obtaining a plurality of preset models, inputting the target prompt word and the transaction certificate into each preset model to obtain a predicted transaction element output by each preset model; determining the number of predicted transaction elements that are the same as the to-be-evaluated transaction element from a plurality of predicted transaction elements corresponding to a plurality of preset models, and taking the number as an evaluation result of the to-be-evaluated transaction element; wherein the evaluation result meets a preset condition, including that the number is greater than a preset threshold.

[0013] A second aspect of the present application provides a transaction certification device, comprising: a first determination module configured to determine a category of a transaction certificate according to transaction information of the transaction certificate in response to obtaining the transaction certificate, the transaction certificate being sent by a transaction object, the transaction information including at least one of a transaction object and a transaction type; a second determination module configured to determine a target prompt word from a preset prompt word set based on the category of the transaction certificate; an acquisition module configured to acquire a target transaction element in the transaction certificate based on the target prompt word using a large model; and a first processing module configured to match the target transaction element with a preset candidate transaction element, and perform transaction certification on the transaction certificate according to a matching result.

[0014] A third aspect of the present application provides an electronic device, comprising: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0015] The fourth aspect of the present application also provides a computer-readable storage medium, which stores a computer program or instructions, and the computer program or instructions are executed by a processor to implement the steps of the above method.

[0016] The fifth aspect of the present application also provides a computer program product, which includes a computer program or instructions, and the computer program or instructions are executed by a processor to implement the steps of the above method.

[0017] According to the embodiments of the present application, by determining the certificate category according to the transaction information in the transaction certificate, and selecting the target prompt word from the preset prompt word set based on the category, the target transaction element can be obtained from the transaction certificate by using the large model and the target prompt word, the element extraction manner can be adaptively adjusted for transaction certificates of different templates, the accuracy of transaction element identification and matching is improved, the workload of manual checking is reduced, the efficiency of transaction certification is improved, and the operation risk caused by transaction element identification error is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and other objects, features and advantages of the present application will become more apparent from the following description when taken in conjunction with the accompanying drawings, in which:

[0019] Figure 1 An application scenario diagram of a transaction certification method, device, equipment, medium and program product according to embodiments of the present application is schematically shown;

[0020] Figure 2 A flowchart of a transaction certification method according to embodiments of the present application is schematically shown;

[0021] Figure 3 A structural block diagram of a transaction certification device according to embodiments of the present application is schematically shown; and

[0022] Figure 4 A block diagram of an electronic device suitable for implementing a transaction certification method according to embodiments of the present application is schematically shown. DETAILED DESCRIPTION

[0023] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present application. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to one skilled in the art that one or more embodiments can be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques have been omitted to avoid unnecessarily obscuring the concept of the present application.

[0024] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. As used herein, the terms "comprises", "comprising", "includes", "including" and the like are to be interpreted open-ended and do not exclude the presence of one or more other features, steps, operations, and / or components.

[0025] All terms used herein including technical and scientific terms have the meanings commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein are to be interpreted as having a meaning that is consistent with the understanding of those terms by those skilled in the art, and should not be interpreted in an idealized or overly formal sense.

[0026] In situations where similar terminology is used, such as "at least one of A, B, and C is used in situations where at least one of the elements A, B, and C are used, in general, it should be understood that such terminology is to be interpreted as the meaning commonly understood by a person of ordinary skill in the art (e.g., "a system having at least one of A, B, and C should be interpreted to include but not be limited to a system that has A alone, a system that has B alone, a system that has C alone, a system that has both A and B, a system that has both A and C, a system that has both B and C, and / or a system that has all of A, B, and C, etc.).

[0027] The following terms are explained and described as follows.

[0028] Transaction verification: a transaction verification document is composed based on key transaction elements, and the transaction parties verify whether the transaction elements in the transaction verification are consistent through the process of checking.

[0029] Optical Character Recognition (OCR): a technology that recognizes and converts the text (printed / handwritten) in pictures / scanned copies into computer editable and searchable text data. The core is "reading".

[0030] Natural Language Processing (NLP): a branch of artificial intelligence that enables computers to understand, interpret, and generate human language (such as text, speech), and realize human-computer language interaction (such as translation, chat robots, sentiment analysis).

[0031] Robotic Process Automation (RPA): using software robots to automatically perform computer operation tasks based on rules and high repetition (such as data entry, form filling), simulating human user interface, and improving efficiency.

[0032] It should be noted that the transaction verification method, device, equipment, medium and program product determined by the present application can be used in the field of artificial intelligence technology and the field of financial technology, and can also be used in various fields other than the field of artificial intelligence technology and the field of financial technology. The application field of the transaction verification method, device, equipment, medium and program product provided by the embodiments of the present application is not limited.

[0033] In the technical solutions of the present application, the user information (including but not limited to user personal information, user image information, user equipment information such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved are information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards, necessary security measures are taken, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.

[0034] In the scenario of using personal information for automated decision-making, the transaction certificate method, device and system provided by the embodiments of the present application all provide corresponding operation portal for the user to choose to agree or refuse the automated decision-making result; if the user chooses to refuse, the expert decision-making process is entered. The expression "expert decision-making" here refers to the decision-making activities of personnel who are engaged in a certain field of work, have special experience, knowledge and skills, and reach a certain professional level.

[0035] Figure 1 The application scenario diagram of the transaction verification method, device, equipment, medium and program product according to the embodiments of the present application is schematically shown.

[0036] As shown in Figure 1 The application scenario 100 according to the embodiment can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104 and a server 105. The network 104 is used as a medium to provide a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0037] The user can use the first terminal device 101, the second terminal device 102, the third terminal device 103 to interact with the server 105 through 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, the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as an example).

[0038] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, and the like.

[0039] The server 105 can be a server providing various services, for example, a background management server supporting a website browsed by a user using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (only as an example). The background management server can perform analysis and the like on received user requests and the like, and feed back a processing result (for example, a webpage, information, or data, or the like, obtained or generated according to a user request) to a terminal device.

[0040] It should be noted that the transaction certification method provided by the embodiments of the present application can generally be executed by the server 105. Correspondingly, the transaction certification apparatus provided by the embodiments of the present application can generally be arranged in the server 105. The transaction certification method provided by the embodiments of the present application can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Correspondingly, the transaction certification apparatus provided by the embodiments of the present application can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0041] It should be understood that the number of terminal devices, networks, and servers in the scenario described in the embodiments of the present application is only illustrative. According to the needs of implementation, there can be any number of terminal devices, networks, and servers. Figure 1

[0042] The transaction certification method according to the embodiments of the present application will be described in detail below based on the scenario described above. Figure 1 Figure 2 The transaction certification method according to the embodiments of the present application will be described in detail below based on the scenario described above.

[0043] Figure 2 A flowchart of the transaction certification method according to the embodiments of the present application is schematically shown.

[0044] As shown in Figure 2 The transaction certification method of this embodiment includes operations S210-S240, which can be executed by a server.

[0045] In operation S210, in response to obtaining a transaction certification document, a category of the transaction certification document is determined according to transaction information of the transaction certification document, the transaction certification document being sent by a transaction object, the transaction information including at least one of a transaction object and a transaction type. ​​

[0046] In the embodiments of the present application, the transaction confirmation can refer to a confirmation document generated by one or both parties after a financial transaction is completed, which is used to specify the transaction object, transaction type, price, term and other key transaction elements in writing.

[0047] For example, the mailbox can be monitored, and whether the received email is for transaction confirmation can be analyzed by analyzing the email header or email body. If it is determined that the email is a transaction confirmation email, the attachment is downloaded to the internal system, and the transaction object, transaction type and other transaction information are parsed from the email header and email body. According to the mapping relationship between the transaction information and the category, the category of the transaction confirmation is automatically determined. It should be noted that the process of monitoring the mailbox, judging the email and downloading the attachment can be realized by the way of robotic process automation.

[0048] For another example, the electronic file uploaded by the transaction object can be obtained through the file upload interface. The file upload interface, for example, is a docking interface opened by the transaction processing system to the transaction object, which can be directly called and used by the transaction object to upload the transaction confirmation. During the process of calling the interface, the request information transmitted can carry fields such as institution code and product type, according to which the transaction object and transaction type and other transaction information can be determined, and then according to the mapping relationship between the transaction information and the category, the category of the transaction confirmation is automatically determined.

[0049] In operation S220, the target prompt word is determined from the preset prompt word set based on the category of the transaction confirmation.

[0050] In the embodiments of the present application, the preset prompt word set can refer to a set of text instruction templates for driving the large model to parse the transaction confirmation, which are configured and stored in advance for different categories of transaction confirmation. The template can specify the transaction element name to be extracted, the output format and the matters needing attention, etc.

[0051] Specifically, the prompt word corresponding to the category in the prompt word set can be determined as the target prompt word according to the transaction category. For example, for foreign exchange forward confirmation, the target prompt word can limit the need to extract elements such as currency pair, nominal principal, exchange rate, interest start date, expiration date, etc. For another example, for different transaction objects, the target prompt word can include the template used by the transaction object, and include which areas in the template can extract the transaction elements.

[0052] In operation S230, the target transaction element in the transaction confirmation is obtained based on the target prompt word using the large model.

[0053] In the embodiments of the present application, the large model can refer to a natural language processing model trained based on massive text data, which can understand and generate natural language text, and perform information extraction, question answering and other tasks according to the input prompt word. The target transaction element can refer to the key business field extracted from the transaction confirmation letter by the large model, which can include transaction amount, transaction date, transaction direction and the like, and it should be noted that the target transaction element extracted can be different for different types of transaction confirmation letters.

[0054] For example, for an insurance transaction confirmation letter, the target transaction element can include contract number, insurance product name, transaction amount, payment method, payment period, etc. For example, for a foreign exchange forward confirmation letter, the target transaction element can include transaction currency pair, nominal principal, exchange rate, interest start date, expiration date, etc.

[0055] In operation S240, the target transaction element is matched with the preset candidate transaction element, and the transaction confirmation is performed on the transaction confirmation letter according to the matching result.

[0056] In the embodiments of the present application, the candidate transaction element can be a plurality of sets of transaction elements extracted from the transaction records in the bank internal transaction system, risk control system or order management system, and each set of transaction elements can correspond to a transaction record. Matching the target transaction element with the preset candidate transaction element can refer to determining whether there is a candidate transaction element in the plurality of sets of candidate transaction elements that is completely consistent with the target transaction element.

[0057] For example, if the target transaction element is completely consistent with the candidate transaction element of a certain transaction record, the matching result can be a successful match, that is, the transaction confirmation letter is consistent with the internal transaction record, and if the target transaction element is not completely consistent with the candidate transaction element of a certain transaction record, the content inconsistent with the candidate transaction element in the transaction confirmation letter can be confirmed by manual method, and returned to the transaction object by email.

[0058] Through the transaction confirmation method of the embodiments of the present application, the type of the transaction confirmation letter is determined according to the transaction information in the transaction confirmation letter, and the target prompt word is selected from the preset prompt word set based on the type. The large model and the target prompt word can be used to obtain the target transaction element from the transaction confirmation letter, which can adaptively adjust the element extraction method for different templates of transaction confirmation letters, improve the accuracy of transaction element identification and matching, reduce the workload of manual checking, improve the efficiency of transaction confirmation, and reduce the operational risk caused by incorrect transaction element identification.

[0059] According to an embodiment of the present application, in response to obtaining the transaction confirmation document, determining the category of the transaction confirmation document according to the transaction information of the transaction confirmation document comprises: in response to obtaining the transaction confirmation document from the target mailbox, determining attribute information of the target mailbox, the attribute information comprising at least one of a sender domain name, a mailbox subject, an attachment type, and a format feature of the transaction confirmation document; extracting at least one of a transaction object and a transaction type from the attribute information as the transaction information of the transaction confirmation document; and processing the transaction information to obtain the category of the transaction confirmation document.

[0060] In the embodiment of the present application, the target mailbox can be a designated business mailbox account for receiving transaction confirmation document mails sent by a customer or a transaction counterparty. When a mail containing a transaction confirmation document is detected and captured from the target mailbox, the attribute information of the mail is automatically read and parsed, such as the domain name of the mail sender, the mail title text, the type of the attached file, and the format feature of the confirmation document in the attachment. The format feature can include the position of the header, the layout of the table, the arrangement of the field title, the font style, the position of the footer signature area, and the like.

[0061] Based on the mail sender domain name, the mail subject, the attachment type, and the format feature, the attribute information is parsed to identify at least one of the transaction object corresponding to the mail and the transaction type involved in the mail, and the identification result is taken as the transaction information of the transaction confirmation document.

[0062] Processing the transaction information can mean classifying the transaction information based on a classification rule or matching the transaction information with a preconfigured category to determine the category of the transaction confirmation document.

[0063] Through the transaction confirmation method of the embodiment of the present application, the transaction object and the transaction type of the transaction confirmation document can be automatically and accurately determined without relying on manual identification, thereby realizing intelligent classification of the confirmation document, improving the accuracy and processing efficiency of the classification, and reducing the risk of manual participation and misjudgment.

[0064] According to an embodiment of the present application, processing the transaction information to obtain the category of the transaction confirmation document comprises: obtaining a plurality of category information in a preset category set, the category information comprising at least one of a transaction object and a transaction type; determining category information matching the transaction information from the plurality of category information as the category of the transaction confirmation document; or calculating the semantic similarity between the transaction information and each of the plurality of category information, and determining the category information with a semantic similarity higher than a preset similarity threshold in the plurality of category information as the category of the transaction confirmation document.

[0065] In the embodiments of the present application, the preset category set can be understood as a preconfigured transaction certificate classification dictionary, wherein each category information at least includes a transaction object transaction type corresponding to the category, for example, the category information is an insurance transaction of institution A and the category information is a securities transaction of institution B, and further can include more detailed contents such as a product name keyword commonly used in the certificate, for enhancing the distinguishing degree.

[0066] Specifically, the transaction certificate templates used for different transaction objects and transaction types can be clustered, and the target of clustering is that the template similarity of the transaction certificates used in each category is greater than a threshold, for example, the transaction certificate templates used for transaction type a by institution A and institution B are basically consistent, then category 1 can be used for institution A or institution B and transaction type a, and the data set used for clustering can be transaction information and corresponding transaction certificate templates, or mail attribute information and corresponding transaction certificate templates.

[0067] In an embodiment, the transaction information can be compared with each category information in the category set to determine the category information matched with the transaction information, and the category information is used as the category of the target transaction certificate. The matching can be exact matching, for example, the transaction object is completely consistent and the transaction type is consistent, or can be fuzzy matching based on rules, for example, the transaction types are allowed to be merged within the same category. When there are multiple category information that simultaneously satisfy the matching condition, one optimal category information can be selected as the final category according to a preset priority, a matching field weight or a time configuration strategy.

[0068] In another embodiment, the transaction information can be processed in a semantic similarity manner. Specifically, the transaction information and the multiple category information can be respectively represented as text vectors that can be used for semantic calculation, for example, a corresponding semantic representation is generated by a pre-trained language model, a word vector or a sentence vector, and then the semantic similarity between the transaction information and each category information is calculated, for example, a cosine similarity or other similarity measurement function is used to obtain the similarity score of the transaction information and each category information. The category information with a semantic similarity higher than a preset similarity threshold in the multiple category information is determined as the category of the transaction certificate.

[0069] Through the transaction certificate identification method of the embodiments of the present application, the preset category set and the rule matching or the semantic similarity calculation can be used to realize automatic identification of the transaction certificate category, ensure accurate matching, have a certain fault tolerance, improve the classification accuracy in the multi-template scenario, and further improve the automation degree and processing efficiency of the transaction certificate business.

[0070] According to an embodiment of the present application, the prompt word set stores a plurality of candidate categories and candidate prompt words corresponding to each candidate category; and the target prompt word is determined from the preset prompt word set based on the category of the transaction certificate, including: determining a candidate category matching the category of the transaction certificate from the plurality of candidate categories, and taking the candidate category as the target category; and determining a candidate prompt word corresponding to the target category in the prompt word set, and taking the candidate prompt word as the target prompt word.

[0071] In the embodiments of the present application, the prompt word set can store a plurality of candidate categories and candidate prompt words corresponding to each candidate category. After determining the category of the transaction certificate, a candidate category matching the category of the transaction certificate can be determined from the plurality of candidate categories in the prompt word set based on the category of the transaction certificate, for example, the prompt word set includes category 1 and prompt word 1 corresponding to category 1, category 2 and prompt word 2 corresponding to category 2. If the category of the transaction certificate is category 1, category 1 can be taken as the target category and prompt word 2 can be taken as the target prompt word.

[0072] For example, the target prompt word can be: you are a financial expert, need to do matching certification work, master the ability to analyze fields, have rich professional knowledge of financial transaction market. Have the ability of fuzzy processing, context understanding, format standardization and exception detection, fuzzy processing is the ability to handle spelling errors, context understanding is to understand the implicit information in the query, format standardization is to unify the data information into a standardized format, and exception detection is to identify abnormal and unresolvable information. The rules for extracting transaction elements include: only processing explicitly stated information, not making any assumptions, maintaining the independence of each piece of information, not making any additional explanations of the extracted information, and strictly limiting within the scope of the query text. The target of extracting transaction elements includes: reading the query text, locating the key information segment, parsing only the following field content from the recognized text input by the user, without outputting redundant fields and content, verifying the accuracy of each piece of information, parsing various formats of time expressions and standardizing them, and directly outputting empty parameters if there is no key information. Notes: date is uniformly formatted, transaction direction only includes buy and sell, actual transaction direction to us is inferred according to the subject of the document type, can automatically identify misplacement of character recognition, intelligently adjust, and output in json format.

[0073] The transaction certification method of the embodiments of the present application preconfigures corresponding candidate prompt words for transaction certificates of different categories, and automatically selects the target prompt word after determining the category of the certificate, which can avoid the problem of inaccurate recognition caused by using a single general prompt word, so that the large model can perform differential analysis for different templates of different transaction objects, and improve the accuracy of element extraction.

[0074] According to an embodiment of the present application, the obtaining of the target transaction element in the transaction certificate based on the target prompt word using the large model comprises: performing layout analysis and character recognition on the transaction certificate respectively to obtain multi-modal features of the transaction certificate, the multi-modal features comprising text content and layout features, the layout features comprising at least one of position information of each text content in the transaction certificate, table structure, and visual layout features; and obtaining the target transaction element in the transaction certificate based on the target prompt word and the multi-modal features using the large model.

[0075] In the embodiments of the present application, the layout analysis can be to analyze the visual structure of the transaction certificate page, automatically split the whole page document into several layout areas, such as a title area, a text area, a table area, a signature area, etc., and identify the coordinate position of each text on the page, the belonging paragraph or table cell, the reading order, the font size, and the bold / underline layout attributes. The character recognition can refer to converting the scanned image or non-editable content in the transaction certificate into a readable text sequence using optical character recognition technology, and identifying the specific content of each character, word, or text line.

[0076] The layout analysis and character recognition on the transaction certificate can obtain the multi-modal features of the transaction certificate, the text content being the semantic information in the transaction certificate, such as the transaction counterpart, the transaction amount, the settlement date, and the like. The layout features are the organization mode and structural relationship of the text content in the page, such as whether the text is a table header, the field is located in the central area below the title, and the like.

[0077] For example, after receiving the target prompt word and the multi-modal features, the large model can understand the semantic information in the certificate by using the text content, identify which text fragments are related to the required transaction element, and on the other hand, understand the document structure by using the layout features, such as judging whether a certain value is a transaction amount or a settlement amount according to the spatial correspondence relationship between the table header cell and the data cell, and judging the meaning of the field according to its row and column position in the table. Finally, the large model locates and extracts the required target transaction element under the joint action of semantic understanding and layout structure understanding.

[0078] Through the transaction certificate method of the embodiments of the present application, the layout analysis and character recognition results are combined with the large model, the multi-modal input composed of the text content and the layout features is used to guide the large model to perceive the table structure and visual layout of the certificate while understanding the semantics, so that the target transaction element can be accurately located and extracted even in the case of complex templates, multi-column layout, and ambiguous field adjacency relationship, and the accuracy of transaction element extraction is improved.

[0079] According to an embodiment of the present application, the obtaining of the target transaction element in the transaction certificate based on the target prompt word using the large model comprises: obtaining a plurality of to-be-evaluated transaction elements in the transaction certificate based on the target prompt word using the large model; performing verification on each to-be-evaluated transaction element based on a plurality of preset verification strategies, determining the confidence of each to-be-evaluated transaction element based on the verification result; and processing the plurality of to-be-evaluated transaction elements based on the confidence to obtain the target transaction element.

[0080] In the embodiments of the present application, the to-be-evaluated transaction element can refer to an unverified transaction element output by the large model based on the target prompt word, and the verification strategy can include at least one of integrity verification, rule verification, and semantic consistency. The integrity verification is used to check whether a field is missing or incomplete, for example, whether it is empty, whether the currency is missing, whether the date appears in pairs, etc. The rule verification is used to check the format and business rule constraints, for example, whether the amount is a positive number, whether the date format is correct, whether the start date is earlier than the expiration date, whether the interest rate falls within a reasonable range, whether the business logic relationship between different fields is met, etc. The semantic consistency verification is used to check whether the transaction element is consistent with the original text context from the semantic perspective, for example, whether the extracted transaction object actually appears in the certificate header or signature, and for example, when the amount appears in multiple places in the transaction certificate, whether the amount is related to the current transaction or the historical balance.

[0081] The verification result can refer to the comprehensive judgment information obtained after applying the above-mentioned plurality of verification strategies for each to-be-evaluated transaction element, which usually includes whether each item of verification is passed, the corresponding error type if it is not passed, etc.

[0082] Based on these verification results, the confidence can be determined, for example, when a certain element passes all verifications, its confidence can be 1, indicating that the element is accurately extracted, and if a key business rule verification fails or there is a serious semantic conflict, the confidence will be relatively low, indicating that there may be a problem in the extraction of the element. Specifically, different weights can be configured for different verification strategies, and the confidence can be calculated through weighted processing.

[0083] The to-be-evaluated element can be screened based on the confidence. For example, the transaction element with a confidence higher than a confidence threshold is directly confirmed as the target transaction element; the transaction element with a confidence lower than the confidence threshold is excluded, marked as abnormal, or handed over to manual review, and the transaction element is modified or determined whether it can be used as the target transaction element according to the review result.

[0084] The transaction verification method of the embodiments of the present application can perform multi-strategy checking on the to-be-evaluated transaction elements, verify and confidence assess each transaction element, automatically eliminate or mark the incorrect, incomplete or unclear semantic fields, improve the accuracy and reliability of the target transaction elements, and reduce the risk of misjudgment.

[0085] According to the embodiments of the present application, the plurality of to-be-evaluated transaction elements are processed based on the confidence to obtain the target transaction element, which includes performing the following operations on any to-be-evaluated transaction element in the plurality of to-be-evaluated transaction elements: in response to the confidence of the to-be-evaluated transaction element being greater than or equal to a first threshold, taking the to-be-evaluated transaction element as the target transaction element; in response to the confidence of the to-be-evaluated transaction element being less than the first threshold and greater than a second threshold, taking the to-be-evaluated transaction element as the target transaction element in a case that confirmation information for the to-be-evaluated transaction element is obtained; wherein the second threshold is less than the first threshold; in response to the confidence of the to-be-evaluated transaction element being less than or equal to the second threshold, evaluating the to-be-evaluated transaction element based on the target prompt word and the transaction verification document, and taking the to-be-evaluated transaction element as the target transaction element in a case that the evaluation result meets a predetermined condition.

[0086] In the embodiments of the present application, the first threshold and the second threshold can be preset, and the first threshold is greater than the second threshold. The confidence represents the probability that the transaction element is correctly extracted, and can also be understood as the credibility that the transaction element is correctly extracted.

[0087] If the confidence of the to-be-evaluated transaction element is greater than or equal to the first threshold, it means that the confidence of the to-be-evaluated transaction element is high, and the to-be-evaluated transaction element can be directly taken as the target transaction element.

[0088] If the confidence of the to-be-evaluated transaction element is less than the first threshold and greater than the second threshold, it means that the confidence of the to-be-evaluated transaction element is general. The to-be-evaluated transaction element can be further confirmed, for example, sent to an artificial for confirmation. When the confirmation information is received to ensure that the transaction element is correctly identified, it can be determined as the target transaction element, otherwise it can be eliminated or modified by an artificial.

[0089] If the confidence level of the transaction element to be evaluated is less than or equal to the second threshold, it indicates that the confidence level of the transaction element is relatively low. Based on the target prompt words and the original content of the transaction confirmation, the transaction element can be re-evaluated. For example, the large model can be called again, and more stringent or targeted prompt words can be used to re-extract or verify the field. Alternatively, the model can be guided to re-evaluate the field's matching degree with the context and its consistency with other fields. After obtaining the evaluation results, if the evaluation results meet predetermined conditions, the transaction element to be evaluated is determined as the target transaction element. If the predetermined conditions are not met, it can be removed or marked as an anomaly. Predetermined conditions include, for example, that the re-parsed results are consistent with the original results and that internal and external rule verifications pass, or that the recalculated confidence level is increased to a certain acceptable range.

[0090] The transaction verification method of the embodiments of this application sets a first threshold and a second threshold and adopts differentiated processing for different confidence intervals, realizing a hierarchical decision-making mechanism. It can achieve fully automatic processing in high confidence scenarios, reducing human intervention, and effectively screen and correct medium and low confidence results, thereby improving the accuracy of target transaction element extraction.

[0091] According to an embodiment of this application, before evaluating the transaction element to be evaluated based on the target prompt words and the transaction confirmation certificate: determine the location information of the transaction element to be evaluated in the transaction confirmation certificate; expand the target prompt words based on the location information.

[0092] In this embodiment of the application, the location information may be information that identifies the page position and structural position of the transaction element to be evaluated in the transaction confirmation document. Before evaluating the transaction element to be evaluated, the page position and structural position of the element in the confirmation document can be located first, and then this location-related information can be converted into text or structured description and attached to the original target prompt words to form expanded prompt words.

[0093] For example, the transaction element to be evaluated is a repurchase rate of 1.95%, located in the second row and third column of the repurchase terms table on page 1, within the other cells of the same row. The first column shows the start date: 2024-01-01, and the second column shows the maturity date: 2024-04-01. The table title is "Clause XXX". The supplementary information could include: This candidate rate of 1.95% is located in the second row and third column of the "Clause XXX" table on page 1, corresponding to the start date 2024-01-01 and maturity date 2024-04-01 in the same row. Please determine whether this value is the repurchase rate for this transaction and, if necessary, re-extract the correct rate from this area.

[0094] By the transaction verification method of the embodiment of the present application, the target prompt word is expanded with position information before evaluation, so that the large model can perceive information such as the page position of the field, the table row and column structure, and the surrounding context when judging the to-be-evaluated transaction element, reduces the confusion caused by the same value appearing in different positions, improves the accuracy and error correction capability of the secondary evaluation of the large model, reduces misjudgment and omissions, and ensures the accuracy of the target transaction element.

[0095] According to the embodiment of the present application, the evaluation of the to-be-evaluated transaction element based on the target prompt word and the transaction verification document includes: obtaining a plurality of preset models, inputting the target prompt word and the transaction verification document into each preset model to obtain a predicted transaction element output by each preset model; determining the number of predicted transaction elements that are the same as the to-be-evaluated transaction element among a plurality of predicted transaction elements corresponding to the plurality of preset models, and taking the number as an evaluation result of the to-be-evaluated transaction element; wherein the evaluation result meets a preset condition includes that the number is greater than a preset threshold.

[0096] In the embodiment of the present application, the preset model can be a natural language processing model with different architectures or different parameters, and the predicted transaction element output by each model can be obtained by inputting the target prompt word and the transaction verification document into each preset model, wherein the target prompt word can be an expanded prompt word.

[0097] After obtaining the prediction results of each preset model, the consistency of a plurality of predicted transaction elements is counted for a specific to-be-evaluated transaction element. Taking the transaction amount as an example, the transaction amount field in the output of each preset model can be read in turn, and compared with the value of the current to-be-evaluated transaction element to determine whether they are the same. The same can be strictly the same in value, or can allow format differences, such as whether to separate with a thousandth separator, whether to attach a currency unit symbol, etc. The number of prediction results that are the same as the to-be-evaluated transaction element in all preset models is counted, and the number is obtained. If the number is greater than a preset threshold, it means that the to-be-evaluated transaction element is supported by multiple models and is a reliable to-be-evaluated transaction element.

[0098] By the transaction verification method of the embodiment of the present application, the risk of misjudgment of a single model can be reduced, and the accuracy and stability of the result of the target transaction element can be improved.

[0099] Figure 3 The structure block diagram of the transaction verification device according to the embodiment of the present application is schematically shown.

[0100] As shown in Figure 3 the transaction verification device 300 of the embodiment includes a first determination module 310, a second determination module 320, an acquisition module 330, and a first processing module 340.

[0101] The first determination module 310 is configured to determine the category of the transaction certificate according to transaction information of the transaction certificate in response to obtaining the transaction certificate, the transaction certificate being sent by a transaction object, and the transaction information including at least one of a transaction object and a transaction type. In an embodiment, the first determination module 310 can be configured to perform operation S210 described above, and details are not repeated here.

[0102] The second determination module 320 is configured to determine a target prompt word from a preset prompt word set based on the category of the transaction certificate. In an embodiment, the second determination module 320 can be configured to perform operation S220 described above, and details are not repeated here.

[0103] The acquisition module 330 is configured to acquire a target transaction element in the transaction certificate based on the target prompt word using a large model. In an embodiment, the acquisition module 330 can be configured to perform operation S230 described above, and details are not repeated here.

[0104] The first processing module 340 is configured to match the target transaction element with a preset candidate transaction element, and perform transaction certification on the transaction certificate according to a matching result. In an embodiment, the first processing module 340 can be configured to perform operation S240 described above, and details are not repeated here.

[0105] According to an embodiment of the present application, the first determination module 310 includes an attribute information determination module, an extraction module, and a second processing module. The attribute information determination module is configured to determine attribute information of a target mailbox in response to obtaining a transaction certificate from the target mailbox, the attribute information including at least one of a sender domain name, a mailbox subject, an attachment type, and a layout feature of the transaction certificate; the extraction module is configured to extract at least one of a transaction object and a transaction type from the attribute information as transaction information of the transaction certificate; and the second processing module is configured to process the transaction information to obtain the category of the transaction certificate.

[0106] According to an embodiment of the present application, the second processing module includes a category information acquisition module, a category information determination module, or a calculation module. The category information acquisition module is configured to acquire a plurality of category information in a preset category set, the category information including at least one of a transaction object and a transaction type; the category information determination module is configured to determine category information matched with the transaction information from the plurality of category information as the category of the transaction certificate; and the calculation module is configured to calculate semantic similarity of the transaction information and each category information in the plurality of category information, and determine category information with a semantic similarity higher than a preset similarity threshold value in the plurality of category information as the category of the transaction certificate.

[0107] According to an embodiment of the present application, the prompt word set stores a plurality of candidate categories and candidate prompt words corresponding to each candidate category; the second determination module 320 includes a target category determination module and a target prompt word determination module: the target category determination module is used to determine a candidate category matching the category of the transaction certificate from the plurality of candidate categories, and serve as the target category; the target prompt word determination module is used to determine the candidate prompt word corresponding to the target category in the prompt word set, and serve as the target prompt word.

[0108] According to an embodiment of the present application, the acquisition module 330 includes a third processing module and an acquisition submodule. The third processing module is used to perform layout analysis and character recognition on the transaction certificate respectively, to obtain the multi-modal feature of the transaction certificate, the multi-modal feature including text content and layout features, the layout features including at least one of position information of each text content in the transaction certificate, table structure and visual layout features; the acquisition submodule is used to acquire the target transaction element in the transaction certificate based on the target prompt word and the multi-modal feature using the large model.

[0109] According to an embodiment of the present application, the acquisition module 330 includes a to-be-evaluated transaction element acquisition module, a fourth processing module and a fifth processing module, the to-be-evaluated transaction element acquisition module is used to acquire a plurality of to-be-evaluated transaction elements in the transaction certificate based on the target prompt word using the large model; the fourth processing module is used to verify each to-be-evaluated transaction element based on a plurality of preset verification strategies, and determine the confidence of each to-be-evaluated transaction element based on the verification result; the fifth processing module is used to process the plurality of to-be-evaluated transaction elements based on the confidence, to obtain the target transaction element.

[0110] According to an embodiment of the present application, the fifth processing module includes a first response module, a second response module and a third response module. For any to-be-evaluated transaction element in the plurality of to-be-evaluated transaction elements, the first response module is used to, in response to the confidence of the to-be-evaluated transaction element being greater than or equal to a first threshold, take the to-be-evaluated transaction element as the target transaction element; the second response module is used to, in response to the confidence of the to-be-evaluated transaction element being less than the first threshold and greater than a second threshold, take the to-be-evaluated transaction element as the target transaction element in a case where confirmation information for the to-be-evaluated transaction element is obtained; wherein the second threshold is less than the first threshold; the third response module is used to, in response to the confidence of the to-be-evaluated transaction element being less than or equal to the second threshold, evaluate the to-be-evaluated transaction element based on the target prompt word and the transaction certificate, and take the to-be-evaluated transaction element as the target transaction element in a case where the evaluation result meets a predetermined condition.

[0111] According to an embodiment of the present application, the third response module further comprises a position information determining module and an expansion module. The position information determining module is configured to determine position information of the transaction element to be evaluated in the transaction certificate. The expansion module is configured to expand the target prompt word based on the position information.

[0112] According to an embodiment of the present application, the third response module comprises an input module and a quantity determining module. The input module is configured to obtain a plurality of preset models, input the target prompt word and the transaction certificate into each preset model respectively, and obtain a predicted transaction element output by each preset model. The quantity determining module is configured to determine a quantity of the predicted transaction elements identical to the transaction element to be evaluated in the plurality of predicted transaction elements corresponding to the plurality of preset models, and take the quantity as an evaluation result of the transaction element to be evaluated. The preset condition met by the evaluation result includes that the quantity is greater than a preset threshold.

[0113] According to an embodiment of the present application, any one or more of the first determining module 310, the second determining module 320, the obtaining module 330 and the first processing module 340 can be combined in one module, or any one of them can be split into multiple modules. Alternatively, at least part of the function of one or more of these modules can be combined with at least part of the function of other modules, and implemented in one module. According to an embodiment of the present application, at least one of the first determining module 310, the second determining module 320, the obtaining module 330 and the first processing module 340 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of hardware or firmware that can be integrated or packaged, or implemented in any one of software, hardware and firmware or in a proper combination of any of them. Alternatively, at least one of the first determining module 310, the second determining module 320, the obtaining module 330 and the first processing module 340 can be at least partially implemented as a computer program module which can perform corresponding functions when running.

[0114] Figure 4 A block diagram of an electronic device suitable for implementing the transaction certification method according to an embodiment of the present application is schematically shown.

[0115] As Figure 4As shown, the electronic device 400 according to embodiments of the present application includes a processor 401 which can perform various appropriate actions and processes in accordance with a program stored in a read only memory (ROM) 402 or a program loaded into a random access memory (RAM) 403 from a storage section 408. The processor 401 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chipset, and / or a special purpose microprocessor (e.g., an application specific integrated circuit (ASIC)), and so on. The processor 401 can also include an on-board memory for cache use. The processor 401 can include a single processing unit or multiple processing units for executing different actions of the method processes according to embodiments of the present application.

[0116] In the RAM 403, various programs and data required for the operation of the electronic device 400 are stored. The processor 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. The processor 401 performs various operations of the method processes according to embodiments of the present application by executing the programs in the ROM 402 and / or the RAM 403. Note that the programs can also be stored in one or more memories other than the ROM 402 and the RAM 403. The processor 401 can also perform various operations of the method processes according to embodiments of the present application by executing the programs stored in the one or more memories.

[0117] According to embodiments of the present application, the electronic device 400 can also include an input / output (I / O) interface 405 which is also connected to the bus 404. The electronic device 400 can also include one or more of the following components connected to the input / output (I / O) interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as necessary. A removable medium 411 such as a magnetic disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 410 as necessary, so that a computer program read out therefrom is installed in the storage section 408 as necessary.

[0118] The application further provides a computer readable storage medium, which can be included in the device / apparatus / system described in the above embodiments, or can exist independently without being 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 the application.

[0119] According to the embodiments of the application, the computer readable storage medium can be a non-volatile computer readable storage medium, which can include, but is not limited to, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any appropriate combination thereof. In this application, a computer readable storage medium can be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to the embodiments of the application, the computer readable storage medium can include the ROM 402 and / or the RAM 403 described above, and / or one or more memories other than the ROM 402 and the RAM 403.

[0120] The embodiments of the application also include a computer program product, which includes a computer program containing program codes for executing the methods shown in the flowcharts. When the computer program product is run in a computer system, the program codes are used to make the computer system implement the transaction verification method provided by the embodiments of the application.

[0121] The above functions defined in the system / apparatus of the embodiments of the application are performed when the computer program is executed by the processor 401. According to the embodiments of the application, the system, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0122] In one embodiment, the computer program can rely on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program can also be transmitted, distributed, and downloaded in the form of a signal on a network medium, and be downloaded and installed through the communication part 409, and / or installed from the detachable medium 411. The program codes contained in the computer program can be transmitted by any appropriate network medium, including but not limited to wireless, wired, etc., or any appropriate combination thereof.

[0123] In such embodiments, the computer program can be downloaded and installed from the network via the communication section 409, and / or installed from the removable media 411. When the computer program is executed by the processor 401, the above-described functions defined in the system of the embodiments of the present application are executed. According to the embodiments of the present application, the system, device, apparatus, module, unit, and the like described above can be realized by the computer program modules.

[0124] According to the embodiments of the present application, the program code for executing the computer program provided by the embodiments of the present application can be written in any combination of one or more programming languages, and specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming language, and / or assembly / machine language. The programming language includes, but is not limited to, such as Java, C++, python, "C" language, or similar programming language. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case involving a remote computing device, the remote computing device can be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, connected to the Internet through an Internet service provider).

[0125] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectural, functional, and operational scenarios of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment, or a portion of code that contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in a different order than that shown in the figures. For example, two blocks that are shown in succession can actually be executed substantially concurrently, or they can sometimes be executed in reverse order, depending on the functionality involved. It should also be noted that each block in the flowcharts or block diagrams, and combinations of blocks in the flowcharts or block diagrams, can be implemented by dedicated hardware-based systems that perform the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0126] Those skilled in the art can understand that the features described in various embodiments of the present application can be combined and / or integrated in various combinations, even if such combinations are not explicitly described in the present application. In particular, the features described in various embodiments of the present application can be combined and / or integrated in various combinations without departing from the spirit and teachings of the present application. All such combinations and / or integrations are within the scope of the present application.

Claims

1. A transaction verification method, characterized in that, The method includes: In response to obtaining a transaction confirmation certificate, the category of the transaction confirmation certificate is determined based on the transaction information in the transaction confirmation certificate, wherein the transaction confirmation certificate is sent by a transaction object, and the transaction information includes at least one of the transaction object and the transaction type; Based on the category of the transaction confirmation, target prompt words are determined from a preset set of prompt words; The large model is used to obtain the target transaction elements in the transaction confirmation document based on the target prompt words; The target transaction element is matched with preset candidate transaction elements, and the transaction confirmation certificate is verified based on the matching result.

2. The method according to claim 1, characterized in that, The step of responding to obtaining a transaction confirmation certificate and determining the category of the transaction confirmation certificate based on the transaction information in the transaction confirmation certificate includes: In response to obtaining a transaction confirmation from the target email address, the attribute information of the target email address is determined, the attribute information including at least one of the following: sender domain, email subject, attachment type, and format characteristics of the transaction confirmation. Extract at least one of the transaction object and transaction type from the attribute information to serve as the transaction information of the transaction certificate; The transaction information is processed to obtain the category of the transaction confirmation certificate.

3. The method according to claim 2, characterized in that, The types of transaction confirmation certificates obtained by processing the transaction information include: Obtain multiple category information from a preset category set, wherein the category information includes at least one of transaction object and transaction type; The category information that matches the transaction information is determined from the plurality of category information and used as the category of the transaction confirmation; or Calculate the semantic similarity between the transaction information and each of the multiple categories of information, and determine the category of the transaction certificate as the category of the multiple categories of information whose semantic similarity is higher than a preset similarity threshold.

4. The method according to claim 1, characterized in that, The prompt word set stores multiple candidate categories and candidate prompt words corresponding to each candidate category; The step of determining target prompt words from a preset prompt word set based on the category of the transaction confirmation includes: From the plurality of candidate categories, a candidate category that matches the category of the transaction confirmation is determined and used as the target category; Candidate prompt words corresponding to the target category are determined from the prompt word set and used as the target prompt words.

5. The method according to claim 1, characterized in that, The method of using a large model to obtain the target transaction elements in the transaction confirmation document based on the target prompt words includes: The transaction confirmation is subjected to layout analysis and character recognition to obtain the multimodal features of the transaction confirmation. The multimodal features include text content and layout features. The layout features include at least one of the following: the position information of each text content in the transaction confirmation, table structure, and visual layout features. Based on the target cue words and the multimodal features, the target transaction elements in the transaction confirmation document are obtained using a large model.

6. The method according to claim 1, characterized in that, The method of using a large model to obtain the target transaction elements in the transaction confirmation document based on the target prompt words includes: The large model is used to obtain multiple transaction elements to be evaluated in the transaction confirmation document based on the target prompt words; Each of the transaction elements to be evaluated is verified based on a variety of preset verification strategies, and the confidence level of each of the transaction elements to be evaluated is determined based on the verification results. The target transaction element is obtained by processing the multiple transaction elements to be evaluated based on the confidence level.

7. The method according to claim 6, characterized in that, The process of processing the plurality of transaction elements to be evaluated based on the confidence level to obtain the target transaction element includes performing the following operations on any one of the plurality of transaction elements to be evaluated: In response to the confidence level of the transaction element to be evaluated being greater than or equal to a first threshold, the transaction element to be evaluated is designated as the target transaction element. In response to the fact that the confidence level of the transaction element to be evaluated is less than a first threshold and greater than a second threshold, if confirmation information for the transaction element to be evaluated is obtained, the transaction element to be evaluated is taken as the target transaction element; wherein, the second threshold is less than the first threshold; In response to the confidence level of the transaction element to be evaluated being less than or equal to the second threshold, the transaction element to be evaluated is evaluated based on the target prompt word and the transaction confirmation certificate. If the evaluation result meets predetermined conditions, the transaction element to be evaluated is taken as the target transaction element.

8. The method according to claim 7, characterized in that, The method further includes, before evaluating the transaction elements to be evaluated based on the target prompt and the transaction confirmation: Determine the location information of the transaction element to be evaluated in the transaction confirmation document; The target prompt is expanded based on the location information.

9. The method according to claim 7, characterized in that, The evaluation of the transaction elements to be evaluated based on the target prompt and the transaction confirmation includes: Multiple preset models are obtained, and the target prompt word and the transaction confirmation are respectively input into each preset model to obtain the predicted transaction elements output by each preset model; Among the multiple predicted transaction elements corresponding to the multiple preset models, the number of predicted transaction elements that are the same as the transaction element to be evaluated is determined, and the number is used as the evaluation result of the transaction element to be evaluated; The evaluation result meeting the preset conditions includes the quantity being greater than the preset threshold.

10. A transaction verification device, characterized in that, The device includes: The first determining module is configured to, in response to obtaining a transaction confirmation certificate, determine the category of the transaction confirmation certificate based on the transaction information in the transaction confirmation certificate, wherein the transaction confirmation certificate is sent by a transaction object, and the transaction information includes at least one of the transaction object and the transaction type; The second determining module is used to determine the target prompt word from a preset prompt word set based on the category of the transaction confirmation certificate; The acquisition module is used to obtain the target transaction elements in the transaction confirmation document based on the target prompt words using a large model; The first processing module is used to match the target transaction element with preset candidate transaction elements, and to verify the transaction certificate based on the matching result.

11. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is 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 9.

12. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 9.

13. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 9.

Images