Contract image piece intelligent comparison method and system based on multi-mode analysis and storage medium thereof

Abstract

The present application relates to the technical field of semantic processing and character recognition, and specifically discloses a contract image piece intelligent comparison method and system based on multi-mode analysis and a storage medium thereof. The technical scheme of the present application acquires the image piece of the agreement, the associated loan note number, product code and stage type, performs optical character recognition on the image piece, matches the contract template from the template library according to the product code and stage type, sequentially performs full-text comparison of the recognized text and each template, extracts key information for each loan note number and performs keyword comparison with the customer information stored on the image platform, and further cross-comparison of the key information of different image pieces under the same loan note number, thereby improving the coverage and accuracy of the agreement audit, and reducing the agreement text deviation omission probability and complaint hidden danger.

Description

Technical Field

[0001] This invention relates to the field of semantic processing and character recognition technology, and in particular to a method, system and storage medium for intelligent comparison of contract image documents based on multi-modal analysis. Background Technology

[0002] In financial business collaborations, the platform regularly sends a large number of customer agreement images to financial institutions. These agreements cover multiple business stages such as credit granting and disbursement, and the document types include various materials such as loan contracts, credit authorization letters, and promissory note confirmation information. Due to the large number of agreements and frequent version updates, traditional manual sampling methods are inefficient and have limited coverage, making it difficult to meet daily compliance management needs.

[0003] Existing automated review methods mostly rely on optical character recognition (OCR) technology to extract text from images. However, their comparison logic is relatively simple, typically only capable of simple text matching or fixed keyword retrieval. They cannot simultaneously perform full-text consistency verification, key information accuracy checks, and cross-correlation verification of multiple documents within the same system. When the same product has multiple business models and corresponding contract templates, existing solutions also struggle to achieve automated polling and comparison between templates and image documents.

[0004] At different stages, such as credit granting and disbursement, there is a lack of information consistency verification mechanisms among multiple image documents under the same IOU number. Mismatches in key elements such as names and ID numbers are difficult to detect in a timely manner. These deficiencies make it easy to overlook issues such as discrepancies in agreement text and inconsistencies in key information, increasing the potential for customer complaints and posing challenges to the prudence and compliance of agreement management.

[0005] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method, system, and storage medium for intelligent comparison of contract images based on multi-mode analysis.

[0007] In a first aspect, the present invention provides an intelligent comparison method for contract image documents based on multi-mode analysis. The technical solution of the intelligent comparison method for contract image documents based on multi-mode analysis is as follows:

[0008] Obtain multiple customer agreement image files and the IOU number, product code, and stage type associated with each customer agreement image file, wherein the stage type includes credit granting stage and disbursement stage;

[0009] Optical character recognition is performed on each customer agreement image to obtain the recognition text corresponding to each customer agreement image;

[0010] Based on the product code and the stage type, match at least one contract template from the template library;

[0011] The identified text is compared with each matched contract template in full text. When multiple contract templates are matched, each template is compared in turn until the comparison difference of one template meets the preset threshold or all templates are compared. The full text comparison difference result is then output.

[0012] For each IOU number, key information is extracted from the recognition text corresponding to each customer agreement image under that IOU number. The key information is then compared with the customer information stored in the image platform using keywords. Furthermore, the key information corresponding to different customer agreement image documents under the same IOU number is cross-compared, and the key information comparison results are output.

[0013] Based on the full-text comparison difference results and the key information comparison results, a comparison result is generated for each IOU number.

[0014] Secondly, the present invention provides an intelligent comparison system for contract image documents based on multi-mode analysis. The technical solution of the intelligent comparison system for contract image documents based on multi-mode analysis is as follows:

[0015] The acquisition module is used to acquire multiple customer agreement image files and the IOU number, product code and stage type associated with each customer agreement image file, wherein the stage type includes credit granting stage and disbursement stage;

[0016] The recognition module is used to perform optical character recognition on each customer agreement image to obtain the recognition text corresponding to each customer agreement image.

[0017] The matching module is used to match at least one contract template from the template library based on the product code and the stage type;

[0018] The output module is used to perform a full-text comparison between the identified text and each matched contract template. When multiple contract templates are matched, each template is compared sequentially until the comparison difference of one template meets the preset threshold or all templates have been compared. The full-text comparison difference result is then output.

[0019] The comparison module is used to extract key information from the recognition text corresponding to each customer agreement image under each IOU number, compare the key information with the customer information stored in the image platform, and cross-compare the key information corresponding to different customer agreement image under the same IOU number, and output the key information comparison result.

[0020] The generation module is used to generate a comparison result for each IOU number based on the full-text comparison difference results and the key information comparison results.

[0021] Thirdly, the technical solution for a storage medium provided by the present invention is as follows:

[0022] The storage medium stores instructions that, when read, cause the storage medium to execute the steps of the intelligent comparison method for contract images based on multi-mode analysis as described in this invention.

[0023] The technical solution of this invention obtains customer agreement images and associated IOU numbers, product codes, and stage types. After performing optical character recognition on the images, it matches contract templates from a template library based on the product code and stage type. The recognized text is then compared with each template in turn. Simultaneously, key information is extracted for each IOU number and compared with customer information stored on the image platform using keywords. Furthermore, key information from different images under the same IOU number is cross-compared. This solves the problems of existing automated review methods having a single comparison logic and being unable to simultaneously complete full-text consistency verification and multi-document cross-correlation verification in the same process. It also solves the problems of difficulty in achieving automated polling comparison when multiple business models correspond to multiple sets of contract templates, and the difficulty in detecting the risk of mismatch of key elements in multiple images under the same IOU number. This improves the coverage and accuracy of agreement review, and reduces the probability of agreement text deviations and omissions, as well as the potential for customer complaints.

[0024] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0027] Figure 1 This is a flowchart illustrating an embodiment of the intelligent comparison method for contract images based on multi-modal analysis according to the present invention.

[0028] Figure 2 This is a schematic diagram of an embodiment of the intelligent comparison system for contract images based on multi-mode analysis according to the present invention. Detailed Implementation

[0029] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0030] Figure 1 The diagram illustrates a flowchart of an embodiment of an intelligent comparison method for contract images based on multi-modal analysis provided by the present invention, executed by a control terminal. Figure 1 As shown, the intelligent comparison method for contract images based on multi-modal analysis includes the following steps:

[0031] S1. Obtain multiple customer agreement image files and the IOU number, product code, and stage type associated with each customer agreement image file, wherein the stage type includes credit granting stage and disbursement stage.

[0032] Among them, the customer agreement image document refers to the various agreement documents signed with the customer and received by the financial institution from the cooperation platform in the form of image files; for example, a PDF scanned copy named "Loan Contract_Customer A_20260701.pdf" contains information such as Customer A's name, ID number, and loan amount.

[0033] Here, the loan receipt number refers to a unique identifier for a loan transaction; for example, the loan receipt number corresponding to the loan contract is "JZ20260701001". The product code refers to a unique identifier for a type of financial product; for example, the product code corresponding to this loan is "PROD_LOAN_A", representing a credit loan product from channel A. The stage type refers to the category of the business stage to which the agreement belongs; for example, this customer agreement image belongs to the credit stage, and the stage type is "credit stage".

[0034] The credit granting stage refers to the business stage where financial institutions conduct credit assessments on customers and determine credit limits. For example, the loan agreement signed by customer A when applying for credit is an agreement in the credit granting stage. The disbursement stage refers to the business stage where customers actually withdraw the loan funds after obtaining the credit limit. For example, if customer A subsequently initiates a withdrawal within the credit limit, the corresponding promissory note confirmation document belongs to the disbursement stage.

[0035] S2. Perform optical character recognition on each customer agreement image to obtain the recognition text corresponding to each customer agreement image.

[0036] The text to be recognized refers to the text content extracted from an image using optical character recognition technology. For example, after performing OCR recognition on a PDF scan, the text to be recognized contains characters such as "Borrower: Customer A, ID number: 1101011234, Loan amount: 100,000 yuan".

[0037] S3. Match at least one contract template from the template library based on the product code and the stage type.

[0038] The template library refers to a database used to store and manage various standard contract templates. For example, the template library contains multiple files such as "Channel A Credit Loan Contract Template V3.0" and "Channel A Disbursement Loan Contract Template V2.1". A contract template refers to a pre-drafted standard agreement text used as a comparison benchmark. For example, the contract template matching the product code "PROD_LOAN_A" and the stage type "Credit Stage" is "Channel A Credit Loan Contract Template V3.0", which specifies an interest rate of 5% and a loan term of 12 months.

[0039] S4. Perform a full-text comparison between the identified text and each matched contract template. When multiple contract templates are matched, compare each template in turn until the comparison difference of one template meets the preset threshold or all templates have been compared. Output the full-text comparison difference result.

[0040] The preset threshold refers to the upper limit of the difference rate pre-set in the full-text comparison to determine whether the identified text and the contract template are consistent. For example, the preset threshold is set to 5%, meaning that the comparison is considered successful when the difference rate is less than 5%. The full-text comparison difference result refers to the statistical information of all differences and their degree after comparing the identified text and the contract template character by character. For example, if the comparison finds that the interest rate in the identified text is 4.5% while that in the template is 5%, this difference is output as part of the full-text comparison difference result.

[0041] S5. For each IOU number, extract key information from the recognition text corresponding to each customer agreement image under that IOU number, compare the key information with the customer information stored in the image platform by keywords, and cross-compare the key information corresponding to different customer agreement image documents under the same IOU number, and output the key information comparison result.

[0042] Here, key information refers to the core data items extracted from the identified text for identity verification; for example, the key information extracted from the identified text is "Customer A" and "1101011234". The image platform refers to a system platform used to store customer agreement image files and their associated business data; for example, the image platform stores loan contract image files and records Customer A's customer information as "Customer A" and "1101011234".

[0043] Customer information refers to standard data related to customer identity stored in the image platform; for example, the customer information stored in the image platform includes the customer name "Customer A" and ID number "1101011234". Key information comparison result refers to the consistency determination result obtained after comparing the extracted key information with the customer information; for example, if the key information "Customer A" and "1101011234" are completely consistent with the customer information, then the key information comparison result is "match successful".

[0044] Among them, different customer agreement image documents under the same loan agreement number include personal information authorization letter, credit authorization letter, and loan agreement. The customer name and customer ID number extracted from the personal information authorization letter, the credit authorization letter, and the loan agreement are compared with the customer information stored on the image platform using keywords. The customer name and customer ID number extracted from the personal information authorization letter, the credit authorization letter, and the loan agreement are also matched for equivalent value. If the key information of any image document is inconsistent with the customer information stored on the image platform, or if the key information of different image documents under the same loan agreement number is inconsistent with each other, it is determined that there is a risk of key element mismatch.

[0045] S6. Based on the full-text comparison difference results and the key information comparison results, generate the comparison results for each IOU number.

[0046] The comparison result refers to the final audit conclusion generated for each loan receipt number after comprehensively comparing the differences in the full text and the comparison results of key information. For example, the comparison result for loan receipt number "JZ20260701001" shows that the full text comparison found a difference in interest rate, the key information comparison passed, and the interest rate clause needs to be manually reviewed.

[0047] The technical solution of this embodiment obtains the customer agreement image file and its associated IOU number, product code, and stage type. After performing optical character recognition on the image file, it matches the contract template from the template library based on the product code and stage type. The recognized text is then compared with each template in turn. At the same time, key information is extracted for each IOU number and compared with the customer information stored on the image platform. Key information of different image files under the same IOU number is also cross-compared. This solves the problems of existing automated review methods having a single comparison logic and being unable to complete full-text consistency verification and multi-document cross-correlation verification in the same process. It also solves the problems of difficulty in achieving automated polling comparison when multiple business models correspond to multiple sets of contract templates, and the difficulty in detecting the risk of mismatch of key elements in multiple image files under the same IOU number. This improves the coverage and accuracy of agreement review, and reduces the probability of agreement text deviation and omission and the risk of customer complaints.

[0048] In one alternative approach, the steps of obtaining multiple customer agreement image files and the IOU number, product code, and stage type associated with each customer agreement image file include:

[0049] The system retrieves customer agreement images, loan receipt numbers, product codes, and stage types in batches from the image platform, and groups customer agreement images with the same loan receipt number into the same image group.

[0050] The same image group refers to a collection of multiple customer agreement image documents with the same loan receipt number. For example, the loan receipt number "JZ20260701001" includes three image documents: loan contract, credit authorization letter, and loan receipt confirmation information. These three documents constitute the same image group.

[0051] Among the above-mentioned optional methods, further steps are taken to batch retrieve customer agreement image files and associated IOU numbers, product codes and stage types from the image platform, and automatically group image files with the same IOU number, which reduces the workload of manual sorting and allows subsequent comparisons to be carried out in an orderly manner according to the IOU dimension.

[0052] In one alternative approach, the step of performing optical character recognition on each customer agreement image to obtain the recognized text corresponding to each customer agreement image includes:

[0053] For each client protocol image, image preprocessing, text region detection, character segmentation, and character recognition are performed sequentially, and the recognized text containing character content and character position coordinates is output.

[0054] Herein, the character content refers to the specific glyph or alphanumeric symbol of each character in the recognized text; for example, the character content of the character "户" in the recognized text is the Chinese character "户", and the character content of "A" is the letter "A". The character position coordinate refers to the two-dimensional position information of each character in the original image file; for example, the character "户" in the recognized text is located at the 100th pixel column and the 200th pixel row at the top left corner of the image, and the coordinate is used to locate the source of the character.

[0055] In the above optional manner, image preprocessing, text region detection, character segmentation and character recognition are further performed on the image file in sequence, so as to output a recognized text containing character content and position coordinates, which provides structured data support for subsequent full-text comparison and key information extraction, and reduces the recognition deviation caused by poor image quality.

[0056] In an optional manner, the template library establishes indexes according to product codes and stage types, and each index position stores one or more contract templates. The step of matching at least one contract template from the template library according to the product code and the stage type comprises:

[0057] Querying the template library by taking the product code and the stage type as a combined key value, and extracting all contract templates corresponding to the combined key value.

[0058] Herein, the combined key value refers to a composite key formed by combining the product code and the stage type for template library query; for example, the product code "PROD_LOAN_A" and the stage type "credit granting stage" are combined into a combined key value "PROD_LOAN_A_credit granting", which is used to extract the corresponding contract template from the template library.

[0059] In the above optional manner, indexes are further established in the template library according to product codes and stage types, and corresponding contract templates are extracted through querying with the combined key value, so that the template matching process is more accurate and faster, the inefficient operation of traversing all templates one by one is avoided, and the service scenario where multiple sets of templates exist for the same product at different stages is adapted.

[0060] In an optional manner, the step of performing full-text comparison on the recognized text and each matched contract template comprises:

[0061] Calculate the number of characters that differ from the identified text and the current contract template character by character. Stop the comparison when the number of characters that differ is less than a preset threshold. When multiple contract templates are matched and the number of characters that differ from the current contract template is not less than the preset threshold, switch to the next contract template and continue the comparison until there is a contract template whose number of characters that differ is less than the preset threshold or all contract templates have been compared. Output the number of characters that differ from the contract template at the time the comparison stops as the full-text comparison difference result.

[0062] The number of differing characters refers to the number of characters that are different when comparing the identified text and the contract template character by character in the full-text comparison. For example, if the identified text and the template have 15 different characters in the positions of interest rate, term, etc., then the number of differing characters is 15.

[0063] In the above-mentioned optional methods, the number of character-by-character differences between the identified text and the contract template is further calculated, and the comparison is stopped when the number of differences is lower than a preset threshold, or when multiple templates are matched, the comparison is switched to the next template in sequence to continue the comparison. This allows the comparison process to be dynamically terminated according to the actual degree of difference, avoiding redundant calculations of irrelevant templates and shortening the overall time consumption of full-text comparison.

[0064] In one alternative approach, the step of extracting key information from the recognition text corresponding to each customer agreement image document under each IOU number includes:

[0065] The target field is located in the recognition text according to a preset keyword list, which includes customer name tags and customer ID number tags. The character sequence at the location of the target field is extracted as the key information.

[0066] The target field refers to a specific field area in the identified text that matches a preset list of keywords. For example, by searching according to the keyword list, the area containing "borrower: customer A" in the identified text is the target field, which includes the customer's name.

[0067] Among them, customer name tags refer to keywords used to locate the location of a customer's name in the identified text; for example, the keyword list includes tags such as "borrower," "name," and "customer name," where the tag "borrower" can locate "customer A." Customer ID number tags refer to keywords used to locate the location of a customer's ID number in the identified text; for example, the keyword list includes tags such as "ID number," "document number," and "ID card," where the tag "ID number" can locate "1101011234."

[0068] In the above-mentioned optional methods, the target fields such as customer name and ID number are located in the recognized text according to the preset keyword list, and the character sequences at the corresponding positions are extracted as key information. This makes the extraction process of key elements have clear rules and reduces the risk of information omission caused by changes in field position or differences in expression.

[0069] In one optional approach, the steps of comparing the key information with the customer information stored on the image platform using keywords, and cross-comparing the key information corresponding to different customer agreement image files under the same IOU number, include:

[0070] Each key information corresponding to each customer agreement image file under each IOU number is matched one by one with the customer information stored in the image platform to generate a first matching result for each key information. The key information corresponding to different customer agreement image files under the same IOU number is matched one by one to generate a second matching result across files. The first matching result and the second matching result are merged and output as the key information comparison result.

[0071] The first matching result refers to the single-file matching record obtained by performing an equivalent matching of the key information extracted from each image under each loan agreement number with the customer information stored on the image platform. For example, the key information "Customer A" and "110101********1234" extracted from the loan agreement are successfully matched with the customer information and recorded as the first matching result "match".

[0072] The second matching result refers to the cross-file consistency record obtained by performing equivalent matching on the key information extracted from different images under the same loan agreement number; for example, the "customer A" extracted from the loan agreement is consistent with the "customer A" extracted from the credit authorization letter, and the ID number is also consistent, which is recorded as the second matching result "cross-file consistency".

[0073] In the above-mentioned optional methods, the key information of each image under each IOU number is further matched with the customer information stored on the image platform, and cross-file matching is performed between different image elements under the same IOU number. The first matching result and the second matching result are generated and then merged and output, so that the accuracy of single-file information and the consistency of multi-file information are verified synchronously in the same process.

[0074] In one alternative approach, the full-text comparison difference results include an edit distance-weighted difference rate, which is calculated according to the following formula:

[0075]

[0076] in, This represents the edit distance weighted difference rate. This indicates the number of different characters in the identified text that have undergone character substitutions relative to the current contract template. This indicates the number of missing characters in the identified text compared to the current contract template. This indicates the number of characters that differ from the current contract template in the identified text due to character insertion. This indicates the total number of characters in the current contract template. This represents the character replacement weighting coefficient. This represents the character missing weight coefficient. This represents the character insertion weight coefficient, and ,when When the value is less than a preset threshold, it is determined that the full-text comparison result between the identified text and the current contract template meets the preset consistency condition.

[0077] It should be noted that the formula for the edit distance weighted difference rate is based on the edit distance theory, which deconstructs the character-level differences between the identified text and the contract template into three basic operations: character replacement, character omission, and character insertion. Adjustable weight coefficients are assigned to each of the three operations. This is because in actual contract comparison scenarios, the degree of impact of different types of character differences on the validity of the agreement varies significantly. Replacement may mean that the content of the clause has been modified, omission may mean that key clauses have been omitted, and insertion may mean that additional content has been added. Using differentiated weights can more accurately reflect the actual degree of impact of various differences on the consistency of the contract.

[0078] The function of the above formula is to normalize the multi-dimensional character differences between the identified text and the contract template into a single quantitative indicator in the form of a percentage. By comparing this indicator with a preset threshold, it can automatically determine whether the current identified text and the current contract template meet the preset consistency conditions, thereby providing an objective basis for the early termination decision in the multi-template polling comparison. At the same time, this weighted difference rate contained in the full-text comparison difference result is also used for the comprehensive generation of subsequent comparison results.

[0079] Among the above-mentioned optional methods, by further introducing an edit distance weighted difference rate, weights are assigned to the three difference types of character replacement, missing, and insertion, and the comprehensive difference ratio is calculated. This allows the full-text comparison results to reflect the actual impact of different types of text deviations, avoids misjudging the severity of differences by a single counting method, and improves the rationality of the comparison judgment.

[0080] Figure 2 This diagram illustrates a structural schematic of an embodiment of an intelligent comparison system for contract images based on multi-modal analysis provided by the present invention. Figure 2 As shown, the intelligent comparison system for contract images based on multi-modal analysis includes:

[0081] The acquisition module 201 is used to acquire multiple customer agreement image documents and the IOU number, product code and stage type associated with each customer agreement image document, wherein the stage type includes credit granting stage and disbursement stage;

[0082] The recognition module 202 is used to perform optical character recognition on each customer agreement image to obtain the recognition text corresponding to each customer agreement image.

[0083] Matching module 203 is used to match at least one contract template from the template library based on the product code and the stage type;

[0084] Output module 204 is used to perform a full-text comparison between the identified text and each matched contract template. When multiple contract templates are matched, each template is compared sequentially until the comparison difference of one template meets the preset threshold or all templates are compared. The full-text comparison difference result is then output.

[0085] The comparison module 205 is used to extract key information from the recognition text corresponding to each customer agreement image under each IOU number, compare the key information with the customer information stored in the image platform, cross compare the key information corresponding to different customer agreement image under the same IOU number, and output the key information comparison result.

[0086] The generation module 206 is used to generate a comparison result for each loan receipt number based on the full-text comparison difference result and the key information comparison result.

[0087] Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0088] An embodiment of the present invention provides a storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned intelligent comparison methods for contract image documents based on multi-mode analysis.

[0089] Alternatively, the storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, or an optical data storage device, etc.

[0090] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a storage medium. The processor of an electronic device reads the computer instructions from the storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned intelligent comparison method for contract image documents based on multi-modal analysis.

[0091] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0092] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0093] The storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, the storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0094] The storage medium carries one or more programs, and when the electronic device executes the one or more programs, the electronic device performs the method shown in the above embodiments.

[0095] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0096] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

Claims

1. A method for intelligent comparison of contract image documents based on multi-modal analysis, characterized in that, include: Obtain multiple customer agreement image files and the IOU number, product code, and stage type associated with each customer agreement image file, wherein the stage type includes credit granting stage and disbursement stage; Optical character recognition is performed on each customer agreement image to obtain the recognition text corresponding to each customer agreement image; Based on the product code and the stage type, match at least one contract template from the template library; The identified text is compared with each matched contract template in full text. When multiple contract templates are matched, each template is compared in turn until the comparison difference of one template meets the preset threshold or all templates are compared. The full text comparison difference result is then output. For each IOU number, key information is extracted from the recognition text corresponding to each customer agreement image under that IOU number. The key information is then compared with the customer information stored in the image platform using keywords. Furthermore, the key information corresponding to different customer agreement image documents under the same IOU number is cross-compared, and the key information comparison results are output. Based on the full-text comparison difference results and the key information comparison results, a comparison result is generated for each IOU number.

2. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 1, characterized in that, The steps for obtaining multiple customer agreement image files and the IOU number, product code, and stage type associated with each customer agreement image file include: The system retrieves customer agreement images, loan receipt numbers, product codes, and stage types in batches from the image platform, and groups customer agreement images with the same loan receipt number into the same image group.

3. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 2, characterized in that, The steps of performing optical character recognition on each customer agreement image to obtain the recognized text corresponding to each customer agreement image include: For each client protocol image, image preprocessing, text region detection, character segmentation, and character recognition are performed sequentially, and the recognized text containing character content and character position coordinates is output.

4. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 3, characterized in that, The template library is indexed according to product code and stage type, with each index location storing one or more contract templates. The step of matching at least one contract template from the template library based on the product code and the stage type includes: The template library is queried using the product code and the stage type as a combined key value to extract all contract templates corresponding to the combined key value.

5. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 4, characterized in that, The step of performing a full-text comparison between the identified text and each matched contract template includes: Calculate the number of characters that differ from the identified text and the current contract template character by character. Stop the comparison when the number of characters that differ is less than a preset threshold. When multiple contract templates are matched and the number of characters that differ from the current contract template is not less than the preset threshold, switch to the next contract template and continue the comparison until there is a contract template whose number of characters that differ is less than the preset threshold or all contract templates have been compared. Output the number of characters that differ from the contract template at the time the comparison stops as the full-text comparison difference result.

6. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 1, characterized in that, For each IOU number, the steps for extracting key information from the recognition text corresponding to each customer agreement image document under that IOU number include: The target field is located in the recognition text according to a preset keyword list, which includes customer name tags and customer ID number tags. The character sequence at the location of the target field is extracted as the key information.

7. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 6, characterized in that, The steps of comparing the key information with the customer information stored on the image platform using keywords, and cross-comparing the key information corresponding to different customer agreement image documents under the same IOU number, include: Each key information corresponding to each customer agreement image file under each IOU number is matched one by one with the customer information stored in the image platform to generate a first matching result for each key information. The key information corresponding to different customer agreement image files under the same IOU number is matched one by one to generate a second matching result across files. The first matching result and the second matching result are merged and output as the key information comparison result.

8. The intelligent comparison method for contract image documents based on multi-modal analysis according to claim 5, characterized in that, The full-text comparison results include an edit distance-weighted difference rate, which is calculated according to the following formula: ； in, This represents the edit distance weighted difference rate. This indicates the number of different characters in the identified text that have undergone character substitutions relative to the current contract template. This indicates the number of missing characters in the identified text compared to the current contract template. This indicates the number of characters that differ from the current contract template in the identified text due to character insertion. This indicates the total number of characters in the current contract template. This represents the character replacement weighting coefficient. This represents the character missing weight coefficient. This represents the character insertion weight coefficient, and ,when When the value is less than a preset threshold, it is determined that the full-text comparison result between the identified text and the current contract template meets the preset consistency condition.

9. A smart comparison system for contract image documents based on multi-modal analysis, characterized in that, include: The acquisition module is used to acquire multiple customer agreement image files and the IOU number, product code and stage type associated with each customer agreement image file, wherein the stage type includes credit granting stage and disbursement stage; The recognition module is used to perform optical character recognition on each customer agreement image to obtain the recognition text corresponding to each customer agreement image. The matching module is used to match at least one contract template from the template library based on the product code and the stage type; The output module is used to perform a full-text comparison between the identified text and each matched contract template. When multiple contract templates are matched, each template is compared sequentially until the comparison difference of one template meets the preset threshold or all templates have been compared. The full-text comparison difference result is then output. The comparison module is used to extract key information from the recognition text corresponding to each customer agreement image under each IOU number, compare the key information with the customer information stored in the image platform, and cross-compare the key information corresponding to different customer agreement image under the same IOU number, and output the key information comparison result. The generation module is used to generate a comparison result for each IOU number based on the full-text comparison difference results and the key information comparison results.

10. A storage medium, characterized in that, The storage medium stores at least one computer program, which, when executed by a processor, implements the intelligent comparison method for contract image documents based on multi-mode analysis as described in any one of claims 1 to 8.

Images