Multilingual document verification method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of multilingual document verification, and discloses a multilingual document verification method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the segmentation, recognition and processing of different language text regions of a scanned image through a preset segmentation model, thereby guaranteeing the precise processing of the unique characteristics of each language, and improving the verification precision of the multilingual document. By establishing the term mapping library, not only is the accuracy of conversion among different languages improved, but also ambiguity and confusion caused by language understanding differences are effectively reduced. The method has the beneficial effects that the compliance of the document is ensured, the compliance risk faced by an enterprise is reduced, the document processing efficiency and quality are improved, the document management capability of the enterprise under the globalization background is enhanced, the operation efficiency and compliance of the enterprise are optimized, and powerful support is provided for the development of international services.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multilingual document verification, and in particular to a multilingual document verification method and device, electronic equipment and a storage medium. BACKGROUND

[0002] Under the background of deepening globalization, multilingual document verification is facing increasingly complex challenges. Currently, many document verification systems are still designed based on mainstream languages and general verification methods, lacking sufficient pertinence and adaptability. This limitation makes these systems often appear to be inadequate when dealing with non-common language documents. For example, in the verification of special language texts, errors, omissions, ambiguities and misunderstandings often occur. The differences in grammar, syntax and cultural background of different languages mean that the corresponding processing methods are also different, which increases the complexity of document verification. The existing technology is difficult to effectively process multilingual documents, and the corresponding processing methods of different languages are also different, which not only leads to the difficulty of multilingual document verification, but also increases the compliance risk of multilingual documents, thereby affecting the document compliance and business efficiency of enterprises. SUMMARY

[0003] Therefore, it is necessary to propose a multilingual document verification method, device, electronic equipment and storage medium to solve the existing multilingual document verification problem.

[0004] A multilingual document verification method, the method comprising: segmenting a scanned image of a document to be verified by a preset segmentation model to obtain a plurality of text region images; identifying the language category corresponding to each of the text region images; obtaining a corresponding document processing method based on the language category of each of the text region images; processing the corresponding text region image based on the document processing method to obtain initial text information corresponding to each of the text region images; obtaining a corresponding term mapping library based on the initial text information, and mapping the initial text information using the corresponding term mapping library to obtain target text information; verifying the target text information to obtain a verification result of the document to be verified.

[0005] Further, after the step of identifying the language category corresponding to each of the text region images, the method further comprises: determining whether each of the text region images has at least two language categories; text region images with at least two language categories are recorded as multilingual text region images; A preset separation model is used to segment the multilingual text region images to obtain text region images corresponding to different language types.

[0006] Further, the step of verifying the target text information to obtain the verification result of the document to be verified includes: Based on the target text information, obtain the verification requirements for each area involved in the document to be verified; According to the verification requirements of each region, obtain the corresponding multiple dimensions of information in the target text information; Determine whether each of the aforementioned dimensional information meets the corresponding region verification requirements; The verification result of the document to be verified is generated based on the judgment result.

[0007] Furthermore, after the step of generating the verification result of the document to be verified based on the judgment result, the method further includes: Extract the dimensional information that does not meet the corresponding region verification requirements and mark it as the target dimensional information; Defect information is generated based on the target dimension information and the corresponding area verification requirements; The defect type for obtaining the defect information; The defect type is matched with the correction schemes in the preset correction scheme library to obtain the target correction scheme corresponding to the defect type.

[0008] Furthermore, after the step of identifying the language type corresponding to each of the text region images, the method further includes: Determine whether any of the stated language types have right-oriented layout; The text region image corresponding to the language type with right-oriented layout is denoted as the right-oriented text region image; The right-directed text region image is mirrored using a preset mirroring method to obtain the standard text region image corresponding to the right-directed text region image.

[0009] Furthermore, before the step of verifying the target text information to obtain the verification result of the document to be verified, the method further includes: Obtain the verification dimensions for each preset region; Obtain the corresponding verification standard based on each of the aforementioned verification dimensions; A verification model is generated based on each of the aforementioned verification standards; wherein, the verification model is used to obtain the verification result of the document to be verified after inputting the target text information.

[0010] Furthermore, the step of obtaining the corresponding terminology mapping library based on the initial text information includes: Obtain the target region and document type corresponding to the initial text information; A corresponding terminology mapping library is obtained based on the target region and the document type; wherein, the terminology mapping library is a data mapping library established in advance according to the terms corresponding to each document type in each preset region.

[0011] A multilingual document verification device, the device comprising: The segmentation module is used to segment the scanned image of the document to be verified using a preset segmentation model to obtain multiple text region images; The recognition module is used to identify the language type corresponding to each of the text region images; The acquisition module is used to acquire the corresponding document processing method based on the language type of each of the text region images; The processing module is used to process the corresponding text region images based on the document processing method to obtain the initial text information corresponding to each text region image; The mapping module is used to obtain the corresponding term mapping library based on the initial text information, and to map the initial text information using the corresponding term mapping library to obtain the target text information; The verification module is used to verify the target text information to obtain the verification result of the document to be verified.

[0012] An electronic device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; Identify the language type corresponding to each of the text region images; Based on the language type of each of the text region images, obtain the corresponding document processing method; The document processing method described above is used to process the corresponding text region images to obtain the initial text information corresponding to each text region image. Based on the initial text information, obtain the corresponding term mapping library, and use the corresponding term mapping library to map the initial text information to obtain the target text information; The target text information is validated to obtain the validation result of the document to be validated.

[0013] A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; Identify the language type corresponding to each of the text region images; Based on the language type of each of the text region images, obtain the corresponding document processing method; The document processing method described above is used to process the corresponding text region images to obtain the initial text information corresponding to each text region image. Based on the initial text information, obtain the corresponding term mapping library, and use the corresponding term mapping library to map the initial text information to obtain the target text information; The target text information is validated to obtain the validation result of the document to be validated.

[0014] The beneficial effects of this invention are as follows: By using a preset segmentation model to segment scanned images, different language text regions are identified and processed, ensuring that the unique characteristics of each language are accurately processed. By establishing a terminology mapping library, not only is the accuracy of translation between different languages ​​improved, but ambiguity and confusion caused by differences in language understanding are also effectively reduced. This ensures document compliance, reduces compliance risks faced by enterprises, improves the efficiency and quality of document processing, helps enhance the document management capabilities of enterprises in a globalized context, optimizes their operational efficiency and compliance, and provides strong support for the development of international business. Attached Figure Description

[0015] 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.

[0016] in: Figure 1 This is a diagram illustrating the application environment of a multilingual document verification method in one embodiment. Figure 2 This is a flowchart of a multilingual document verification method in one embodiment; Figure 3 This is a structural block diagram of a multilingual document verification device in one embodiment; Figure 4 This is a structural block diagram of an electronic device in one embodiment. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Figure 1 This is a diagram illustrating a multilingual document verification application environment in one embodiment. (Refer to...) Figure 1 This multilingual document verification method is applied to a multilingual document verification system. The system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet, or laptop. The server 120 can be a standalone server or a server cluster consisting of multiple servers. The terminal 110 is used to acquire scanned images of the document to be verified, and the server 120 is used to generate the verification results for the document.

[0019] like Figure 2 As shown, in one embodiment, a multilingual document verification method is provided. This method can be applied to both terminals and servers; this embodiment uses terminal application as an example. The multilingual document verification method specifically includes the following steps: S1: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; S2: Identify the language type corresponding to each of the text region images; S3: Obtain the corresponding document processing method based on the language type of each of the text region images; S4: Process the corresponding text region images based on the document processing method to obtain the initial text information corresponding to each text region image; S5: Obtain the corresponding term mapping library based on the initial text information, and use the corresponding term mapping library to map the initial text information to obtain the target text information; S6: Verify the target text information to obtain the verification result of the document to be verified.

[0020] As described in step S1 above, the scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images. That is, the paper document to be verified can be scanned using traditional OCR tools to obtain a visual image, which is then divided into multiple smaller text region images. This segmentation operation can be performed using computer vision and deep learning techniques to ensure high-precision region extraction. Specifically, the system preprocesses the input scanned image to eliminate noise and improve image clarity, using techniques such as grayscale conversion and binarization to enhance recognizability. Next, a deep learning segmentation network (such as U-Net or Mask R-CNN) is used to analyze and identify different elements in the image, accurately dividing multiple text regions. These elements can specifically be the features of the text in each preset region.

[0021] As described in step S2 above, the language type corresponding to each of the text region images is identified. For each obtained text region image, language type identification is performed separately. Since different languages ​​have different characteristics, grammatical structures, and processing methods, the system employs machine learning or deep learning algorithms to achieve accurate language recognition. These are typically trained convolutional neural networks (CNNs) or Transformer-based models (such as BERT). First, the system extracts features from each text region, including letter shape, frequency, and character distribution. Then, it compares these features with known language models to obtain the most probable language type. A multilingual classifier can also be used, taking into account factors such as symbols, spelling, and vocabulary during the recognition process. This systematically labels the language type of each text region and provides the necessary basis for selecting subsequent document processing methods.

[0022] As described in step S3 above, corresponding document processing methods are obtained based on the language type of each text region image. Different languages ​​require different processing strategies. Based on the identified language type, a pre-built document processing method database is used to match processing techniques, rules, and standards related to that specific language. For example, for right-oriented languages, it's necessary to handle connected characters and right-oriented layouts. For languages ​​with complex grammar, more emphasis may be placed on analyzing their phonetic characteristics and grammatical structures. For structurally complex characters, stroke reconstruction algorithms may be needed to improve the accuracy of recognizing blurry scanned documents. Furthermore, the processing method for each language may need to be fine-tuned for the differentiated needs of specific industries or fields (such as law, finance, and technology) to ensure that the most appropriate mechanism is used to acquire initial text information for different languages. Specifically, the document processing method can be a method using Optical Character Recognition (OCR) technology, i.e., different OCR algorithms are pre-designed for processing, with each OCR algorithm corresponding to a different document processing method.

[0023] As described in step S4 above, the corresponding text region images are processed based on the document processing method to obtain the initial text information corresponding to each text region image. According to the determined document processing method, each text region image is specifically processed to obtain the initial text information corresponding to each region. Specifically, Optical Character Recognition (OCR) technology can be applied to convert image data into machine-readable text. During this process, the system selects the most suitable OCR algorithm based on the characteristics of different languages. For example, for languages ​​with connected scripts (such as Arabic), a specially designed OCR model can be used to improve recognition accuracy, while for alphabetic languages ​​(such as English), mature general-purpose OCR tools can be used. Furthermore, text quality improvement processing is performed in this stage, including spelling correction, formatting, and grammar correction, to ensure that the generated initial text information is as accurate as possible. The final output initial text information will become the basis for subsequent steps of terminology mapping and verification.

[0024] As described in step S5 above, a corresponding terminology mapping library is obtained based on the initial text information, and the initial text information is mapped using the corresponding terminology mapping library to obtain the target text information. Based on the obtained initial text information, the corresponding terminology mapping library is extracted, and the initial text information is mapped using this mapping library to finally generate the target text information. A terminology mapping library is an important tool to ensure accuracy even in language translation, especially in cross-border or industry-specific documents, where terminology usage not only covers direct translation but also involves industry-specific professional terms. Therefore, the system needs to extract relevant entries from the previously established terminology library according to the identified language type and perform translation and mapping operations for the corresponding language, making the obtained target text information more consistent with the specifications required by the target document. This process can improve the acceptability and compliance of the final document, especially in the context of increasingly stringent external agency review and compliance requirements. Specifically, a comparison matrix of different text types in various regions can be established, and then regional expression differences can be embedded to form a terminology mapping library.

[0025] As described in step S6 above, the target text information is verified to obtain the verification result of the document to be verified. An automated review mechanism can be used to determine whether the target text information complies with relevant laws, regulations, and industry standards. The verification process can cover multiple levels, including the completeness of the content (whether important information is missing), the consistency of terminology (whether specific terms are used consistently in different paragraphs), and the correctness of the grammar (ensuring no spelling or grammatical errors). Machine learning models can also be combined to detect potential compliance risks based on patterns and trends generated from historical data. Furthermore, the system may compare the verification results with the requirements of specific regulations to ensure the compliance and legality of the document.

[0026] In one embodiment, after step S2 of identifying the language type corresponding to the text region image, the method further includes: S301: Determine whether each of the text region images contains at least two languages; S302: Text region images with at least two languages ​​are denoted as multilingual text region images; S303: The multilingual text region image is segmented using a preset separation model to obtain text region images corresponding to different language types.

[0027] As described in step S301 above, it is determined whether each text region image contains at least two languages. Specifically, the system analyzes the language markers of each text region to check for the presence of different languages ​​side-by-side. This is typically achieved through statistical analysis and conditional checks of language types. If only one language is identified in a text region, it is labeled as a monolingual text region; conversely, if other languages ​​related to that language are found, it is labeled as a multilingual text region. Comparison methods using different language features, such as character statistics (e.g., frequency of occurrence of specific characters) and entity recognition (e.g., specific proper nouns), can be used to ensure the accuracy of the determination. Finally, the language type determination results are stored as input for subsequent steps, both for assessing the accuracy of text region labeling and for preparing for subsequent segmentation work.

[0028] As described in step S302 above, text region images identified as having at least two languages ​​in the judgment results are marked and referred to as multilingual text region images. These multilingual text region images are extracted from the entire document and categorized. The information from these regions is integrated into the entire document processing flow to ensure that subsequent operations can be performed specifically. The system may use different colors, highlighted markers, or other visual cues to clearly distinguish these multilingual text regions, facilitating quick identification and judgment by the user on the interface.

[0029] As described in step S303 above, a pre-set separation model is used to accurately segment the text region images labeled as multilingual, with the aim of separating the text region images corresponding to different language types. This segmentation process is necessary because in multilingual text regions, mixed languages ​​may lead to ambiguity of information and reduced accuracy. The system identifies specific text boundaries based on the characteristics of each language type and extracts them separately. This goal can be achieved based on image processing techniques and deep learning models, such as convolutional neural networks (CNNs) or other segmentation techniques, such as U-Net or Mask R-CNN models. These models have been trained to accurately identify and separate various different text elements. After segmentation, the image regions of each language text will be preserved in the newly generated output image, thus ensuring that each language can enter the subsequent text recognition and processing stage independently.

[0030] In one embodiment, step S6, which verifies the target text information to obtain the verification result of the document to be verified, includes: S601: Obtain the verification requirements for each region of the document to be verified based on the target text information; S602: Obtain multiple dimension information corresponding to the target text information according to the verification requirements of each region; S603: Determine whether each of the dimensional information meets the corresponding region verification requirements; S604: Generate the verification result of the document to be verified based on the judgment result.

[0031] As described in step S601 above, based on the target text information, the verification requirements for each area of ​​the document to be verified are obtained. The core of this process is to extract information such as laws, regulations, industry standards, and company policies related to specific areas to ensure the compliance and accuracy of all text content. First, the system parses the target text information and identifies different areas in the document, such as tax clauses, disclaimers, and definitions. Then, for each area, the system retrieves predefined verification requirements. These requirements may include specific formats, content, necessary declarations, the use of key terms, or whether specific legal clauses are accurately cited. Matching can be performed using a knowledge base or regulatory database to obtain the latest compliance requirements, ensuring that the target text information has a basis for verification during the process, making subsequent judgments more scientific and accurate, and avoiding compliance risks caused by process omissions.

[0032] As described in step S602 above, based on the previously acquired verification requirements for each region, the system extracts corresponding multi-dimensional information from the target text information. Specifically, the system analyzes the target text content and identifies important information related to the verification requirements, such as tax type, amount, calculation method, relevant legal clauses, date, and number format. This dimensional information can be automatically extracted using Natural Language Processing (NLP) technology. The system employs techniques such as keyword matching, entity recognition, and relation extraction to complete this process. The extracted dimensional information is then organized into structured data for easier subsequent review and comparison. By acquiring this dimensional information, the system ensures that the content of each region can be mapped to the corresponding verification requirements, thereby laying the foundation for subsequent compliance judgments.

[0033] As described in step S603 above, the system determines whether the information in each dimension meets the corresponding regional verification requirements. First, the system compares the acquired dimensional information, simultaneously verifying whether each piece of information meets the previously defined verification requirements. This can be done through a pre-defined rule engine; if any dimension fails to meet the requirements, the system will mark it. Furthermore, applying machine learning models can improve the intelligence of the judgment. The system can learn to recognize past compliance performance, thereby dynamically adjusting the judgment criteria. The result of the judgment will be a binary classification, i.e., "meets" or "does not meet," and the system can record the specific matching situation of each dimension's information, forming a detailed verification log.

[0034] As described in step S604 above, the final verification result of the document to be verified is generated based on the judgment result. This summarizes which dimensions meet the verification requirements and which do not, and generates a clear verification report. The report typically includes the compliance status of each area, a description of the problem, a risk statement, and recommended actions. This summary not only helps users quickly identify potential problems in the text but also provides necessary evidence for compliance review. The generated verification results can also be recorded in the database as a reference for future improvements. With this result, companies can take necessary corrective measures to ensure that the document ultimately complies with relevant laws, regulations, and internal standards.

[0035] In one embodiment, after step S604 of generating the verification result of the document to be verified based on the judgment result, the method further includes: S6051: Extract the dimensional information that does not meet the corresponding region verification requirements and mark it as the target dimensional information; S6052: Generate defect information based on the target dimension information and the corresponding area verification requirements; S6053: Obtain the defect type of the defect information; S6054: Match the defect type with the correction schemes in the preset correction scheme library to obtain the target correction scheme corresponding to the defect type.

[0036] As described in step S6501 above, the dimension information that does not meet the corresponding region's verification requirements is extracted and marked as target dimension information. This involves processing the identified "non-compliant" results, retrieving relevant specific dimension information such as tax rates, terms and conditions, key dates, or amounts, and then creating a list or table to clearly show each non-compliant item.

[0037] As described in step S6052 above, defect information is generated based on the extracted target dimension information and corresponding regional verification requirements. The key to this process is transforming the specific dimension information that does not meet the requirements into actionable defect descriptions to facilitate subsequent repairs. This process typically includes the following steps: First, the system compares the unmet dimension information with the actual requirements to identify specific discrepancies, such as incorrect tax rates, redundant or missing terms, and formatting irregularities. The system can generate a defect information report, containing details such as a specific defect description, affected areas, the dimension information involved, and the current document status. This defect information helps users quickly identify and understand the problem, and also provides necessary basis for subsequent correction strategies.

[0038] As described in step S6053 above, the system analyzes the generated defect information to determine its defect type. Specifically, based on the description and characteristics in the defect information, the system uses logical judgment, keyword matching, or rule-based classification algorithms to categorize the defect information into specific defect types. For example, defect types may include non-compliant formatting, incorrect content, inconsistent terminology, missing information (detecting missing voucher chains), and locating contradictory amounts. Through this classification process, the system simplifies the workload for subsequent processing while ensuring an efficient processing flow, as different defect types typically have different requirements and handling methods in their correction strategies. Understanding defect types not only helps in quickly locating problems but also lays a good foundation for the standardization and systematization of remediation plans, further improving the efficiency and quality of text processing.

[0039] As described in step S6054 above, the system matches the acquired defect type with the corresponding correction solutions in the pre-set correction solution library to obtain the target correction solution. Specifically, the system uses the defect type as input and retrieves relevant correction measures from the stored correction solution library. These correction solutions cover various methods, such as text formatting adjustments, terminology changes, and necessary legal clause supplementation. During the matching process, the system considers multiple factors, including the user's operation history, best practices, and the results of similar cases, to ensure that the provided correction strategies are targeted and practical. Finally, the matched target correction solution will be fed back to the user, forming a complete defect correction process that enables enterprises to quickly and effectively resolve compliance issues in documents and improve the efficiency and accuracy of document verification. The correction solution library is a pre-set database in which corresponding correction solutions are manually formulated according to the defect type.

[0040] In one embodiment, after step S2 of identifying the language type corresponding to each of the text region images, the method further includes: S311: Determine whether any of the stated language types has a right-oriented layout; S312: The text region image corresponding to the language type with right-oriented layout is denoted as the right-oriented text region image; S313: The right-direction text region image is mirrored using a preset mirror flipping method to obtain the standard text region image corresponding to the right-direction text region image.

[0041] As described in steps S311-S313 above, the system identifies whether any of the identified language types are right-oriented languages. A list of language types can be created and compared with predefined right-oriented language types. Through this comparison, the system can identify whether a right-oriented language appears in the current text area. That is, various languages ​​are pre-stored in the system, and right-oriented languages ​​are marked so that subsequent identification only requires recognizing the characteristics of these languages ​​to determine if they are right-oriented. If at least one right-oriented language is found, the system marks the text area as a right-oriented text area for further processing. A preset mirroring method is used to mirror the identified right-oriented text area image to convert it into a standard text area image, transforming the right-oriented language format into a format that is easier for computers to process and recognize. Since most OCR technologies and text processing systems are primarily designed to handle left-to-right text writing formats, mirroring is necessary to adjust the layout of right-oriented text. The system uses image processing algorithms (such as the OpenCV library) to horizontally flip the right-hand text region image, generating a corresponding standard text region image. Specifically, it imports the OpenCV library and other potentially used libraries (such as NumPy), uses OpenCV's `cv2.imread()` function to read the right-hand text region image, and then uses OpenCV's `cv2.flip()` function to horizontally flip the image. This process not only ensures that the text is correctly defined and read, but also maximizes the accuracy of OCR recognition in subsequent steps. Furthermore, the generated standard text region image after the mirroring operation is compatible with text processing tools in other languages, giving the system greater flexibility and accuracy in multilingual processing. This operation also removes obstacles for subsequent text recognition and verification steps, improving overall processing efficiency and effectiveness. It should be noted that the flipping is only for image preprocessing and does not affect the text semantics; after OCR recognition, the text must be flipped back to its original word order.

[0042] In one embodiment, before step S6 of verifying the target text information to obtain the verification result of the document to be verified, the method further includes: S501: Obtain the verification dimensions of each preset region; S502: Obtain the corresponding verification standard based on each of the aforementioned verification dimensions; S503: Generate a verification model based on each of the verification standards; wherein, the verification model is used to obtain the verification result of the document to be verified after inputting the target text information.

[0043] As described in step S501 above, the verification dimensions for each preset area related to the document to be verified are obtained. Specifically, based on the document's structure and content classification, different preset areas in the document are identified, such as tax clauses, legal statements, and key data. These areas have different verification requirements, therefore they need to be processed separately. The system will extract the verification dimensions corresponding to each area as a unit, such as validity, completeness, consistency, and accuracy. To ensure the comprehensiveness and accuracy of the verification dimensions, the system can refer to resources such as regulations, industry standards, and internal corporate policies. In addition, by obtaining the verification dimensions, the system can lay a clear framework for the subsequent construction of the verification model, ensuring the scientific nature and consistency of the verification process, thereby improving the efficiency of document compliance.

[0044] As described in step S502 above, based on the extracted verification dimensions, the corresponding verification standards are obtained. Specifically, the system utilizes a pre-established verification standard library to extract matching standards by comparing each extracted dimension. For example, for the "accuracy" dimension, it may be necessary to verify whether the numbers and amounts are filled in correctly; while the "completeness" standard may include whether all necessary fields have been filled in. These verification standards can come from internal regulations, relevant laws, or industry standards. Specific standards facilitate subsequent model building and text verification. Obtaining standards not only provides a concrete basis for subsequent verification but also helps users better understand the specific requirements of each region when needed, thereby reducing compliance risks.

[0045] As described in step S503 above, the system generates a verification model based on the acquired verification criteria. This model is used to input target text information and output the verification result of the document to be verified. First, the system designs a suitable model architecture based on the verification criteria set in the previous steps, possibly using a combination of machine learning or rule engine strategies. Specifically, the model receives target text information and analyzes and evaluates it according to the defined verification criteria. The model can be constructed using supervised learning algorithms, such as decision trees, random forests, and support vector machines (SVMs). These algorithms can learn how to determine whether text information meets the verification criteria based on labeled data. In addition, if the document content is complex, the system can also consider using deep learning methods to perform comprehensive analysis through deep neural networks.

[0046] It should be noted that the verification model can be obtained by training a pre-built first neural network model based on a preset sample set. Each sample data in the preset sample set includes text information data and corresponding scores for each dimension. When training the pre-built first neural network model, the text information data from each sample data is used as the input, and the scores for each dimension are used as the output. Through training, the first neural network model can learn the correspondence between all possible text information data and the scores for each dimension. The trained first neural network model is used as the initial verification model, and then the corresponding verification criteria are obtained as parameters for determining whether the verification is successful, thus forming the verification model.

[0047] In one embodiment, step S5, which involves obtaining the corresponding terminology mapping library based on the initial text information, includes: S511: Obtain the target region and document type corresponding to the initial text information; S512: Obtain the corresponding terminology mapping library based on the target region and the document type; wherein, the terminology mapping library is a data mapping library pre-established according to the terms corresponding to each document type in each preset region.

[0048] As described in steps S511-S512 above, the corresponding target regions and document types are extracted from the initial text information. The function and purpose of each specific region in the document are identified. Since different regions typically have different properties and requirements, the corresponding terminology used also varies significantly. The document type is determined; common document types may include declaration forms, audit reports, financial statements, etc. The classification of document types will affect the selection and mapping of terms, as different types of documents may differ in terminology application and expression habits. Based on the obtained target regions and document types, the corresponding terminology mapping library is retrieved. The terminology mapping library is a pre-set dataset that, based on industry standards, regulatory requirements, and domain expertise, meticulously records and maps specific terms used by different document types in each pre-set region. First, the system quickly retrieves relevant terms from the terminology mapping library based on the nature of the target region and the document type.

[0049] Reference Figure 3 The present invention also provides a multilingual document verification device, the device comprising: The segmentation module 902 is used to segment the scanned image of the document to be verified using a preset segmentation model to obtain multiple text region images; The recognition module 904 is used to identify the language type corresponding to each of the text region images; The acquisition module 906 is used to acquire the corresponding document processing method based on the language type of each of the text region images; The processing module 908 is used to process the corresponding text region images based on the document processing method to obtain the initial text information corresponding to each text region image. The mapping module 910 is used to obtain the corresponding term mapping library based on the initial text information, and to map the initial text information using the corresponding term mapping library to obtain the target text information; The verification module 912 is used to verify the target text information to obtain the verification result of the document to be verified.

[0050] In one embodiment, the multilingual document verification device further includes: The first text region image determination module is used to determine whether each of the text region images contains at least two languages. A multilingual text region image labeling module is used to label text region images with at least two languages ​​as multilingual text region images; The multilingual text region image segmentation module is used to segment the multilingual text region image using a preset separation model to obtain text region images corresponding to different language types.

[0051] In one embodiment, the verification module 912 includes: The region verification requirement acquisition submodule is used to acquire the region verification requirements of the document to be verified based on the target text information. The dimension information acquisition submodule is used to acquire multiple dimension information corresponding to the target text information according to the verification requirements of each region. The dimension information judgment submodule is used to determine whether each dimension information meets the corresponding region verification requirements; The verification result generation submodule is used to generate the verification result of the document to be verified based on the judgment result.

[0052] In one embodiment, the verification module 912 further includes: The target dimension information marking submodule is used to extract dimension information that does not meet the corresponding region verification requirements and mark it as target dimension information; The defect information generation submodule is used to generate defect information based on the target dimension information and the corresponding area verification requirements; The defect type acquisition submodule is used to acquire the defect type of the defect information. The correction scheme matching submodule is used to match the defect type with the correction schemes in the preset correction scheme library to obtain the target correction scheme corresponding to the defect type.

[0053] In one embodiment, the multilingual document verification device further includes: The language type determination module is used to determine whether any of the stated language types are right-oriented languages. The right-direction text region image marking module is used to mark the text region images corresponding to languages ​​with right-direction typography as right-direction text region images; The flipping module is used to mirror the right-directed text region image using a preset mirror flipping method to obtain the standard text region image corresponding to the right-directed text region image.

[0054] In one embodiment, the multilingual document verification device further includes: The verification dimension acquisition module is used to acquire the verification dimensions of each preset area; The verification standard acquisition module is used to acquire the corresponding verification standard based on each of the verification dimensions. The verification model generation module is used to generate a verification model based on each of the verification standards; wherein, the verification model is used to obtain the verification result of the document to be verified after inputting the target text information.

[0055] In one embodiment, the mapping module 910 includes: The document type acquisition submodule is used to acquire the target area and document type corresponding to the initial text information; The terminology mapping library acquisition submodule is used to acquire the corresponding terminology mapping library based on the target region and the document type; wherein, the terminology mapping library is a data mapping library pre-established according to the terms corresponding to each document type in each preset region.

[0056] Figure 4 An internal structural diagram of an electronic device in one embodiment is shown. This electronic device can specifically be a terminal or a server, and more specifically, a computer device. Figure 4 As shown, the electronic device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement a multilingual document verification method. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to implement the multilingual document verification method. Those skilled in the art will understand that... Figure 4The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0057] In one embodiment, an electronic device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; Identify the language type corresponding to each of the text region images; Based on the language type of each of the text region images, obtain the corresponding document processing method; The document processing method described above is used to process the corresponding text region images to obtain the initial text information corresponding to each text region image. Based on the initial text information, obtain the corresponding term mapping library, and use the corresponding term mapping library to map the initial text information to obtain the target text information; The target text information is validated to obtain the validation result of the document to be validated.

[0058] By employing a pre-defined segmentation model, scanned images are segmented, and different language text regions are identified and processed. This ensures that the unique characteristics of each language are accurately handled. By establishing a terminology mapping database, not only is the accuracy of translation between different languages ​​improved, but ambiguity and confusion caused by differences in language understanding are also effectively reduced. This ensures document compliance, reduces compliance risks faced by enterprises, improves the efficiency and quality of document processing, helps enhance the document management capabilities of enterprises in a globalized context, optimizes their operational efficiency and compliance, and provides strong support for the development of international business.

[0059] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following steps: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; Identify the language type corresponding to each of the text region images; Based on the language type of each of the text region images, obtain the corresponding document processing method; The document processing method described above is used to process the corresponding text region images to obtain the initial text information corresponding to each text region image. Based on the initial text information, obtain the corresponding term mapping library, and use the corresponding term mapping library to map the initial text information to obtain the target text information; The target text information is validated to obtain the validation result of the document to be validated.

[0060] By employing a pre-defined segmentation model, scanned images are segmented, and different language text regions are identified and processed. This ensures that the unique characteristics of each language are accurately handled. By establishing a terminology mapping database, not only is the accuracy of translation between different languages ​​improved, but ambiguity and confusion caused by differences in language understanding are also effectively reduced. This ensures document compliance, reduces compliance risks faced by enterprises, improves the efficiency and quality of document processing, helps enhance the document management capabilities of enterprises in a globalized context, optimizes their operational efficiency and compliance, and provides strong support for the development of international business.

[0061] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0062] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A multilingual document verification method, characterized in that, The method includes: The scanned image of the document to be verified is segmented using a preset segmentation model to obtain multiple text region images; Identify the language type corresponding to each of the text region images; Based on the language type of each of the text region images, obtain the corresponding document processing method; The document processing method described above is used to process the corresponding text region images to obtain the initial text information corresponding to each text region image. Based on the initial text information, obtain the corresponding term mapping library, and use the corresponding term mapping library to map the initial text information to obtain the target text information; The target text information is validated to obtain the validation result of the document to be validated.

2. The multilingual document verification method according to claim 1, characterized in that, After the step of identifying the language type corresponding to the text region image, the method further includes: Determine whether each of the text region images contains at least two languages; Text region images containing at least two languages ​​are denoted as multilingual text region images; A preset separation model is used to segment the multilingual text region images to obtain text region images corresponding to different language types.

3. The multilingual document verification method according to claim 1, characterized in that, The step of verifying the target text information to obtain the verification result of the document to be verified includes: Based on the target text information, obtain the verification requirements for each area involved in the document to be verified; According to the verification requirements of each region, obtain the corresponding multiple dimensions of information in the target text information; Determine whether each of the aforementioned dimensional information meets the corresponding region verification requirements; The verification result of the document to be verified is generated based on the judgment result.

4. The multilingual document verification method according to claim 3, characterized in that, After the step of generating the verification result of the document to be verified based on the judgment result, the method further includes: Extract the dimensional information that does not meet the corresponding region verification requirements and mark it as the target dimensional information; Defect information is generated based on the target dimension information and the corresponding area verification requirements; The defect type for obtaining the defect information; The defect type is matched with the correction schemes in the preset correction scheme library to obtain the target correction scheme corresponding to the defect type.

5. The multilingual document verification method according to claim 1, characterized in that, After the step of identifying the language type corresponding to each of the text region images, the method further includes: Determine whether any of the stated language types have right-oriented layout; The text region image corresponding to the language type with right-oriented layout is denoted as the right-oriented text region image; The right-directed text region image is mirrored using a preset mirroring method to obtain the standard text region image corresponding to the right-directed text region image.

6. The multilingual document verification method according to claim 1, characterized in that, Before the step of verifying the target text information to obtain the verification result of the document to be verified, the method further includes: Obtain the verification dimensions for each preset region; Obtain the corresponding verification standard based on each of the aforementioned verification dimensions; A verification model is generated based on each of the aforementioned verification standards; wherein, the verification model is used to obtain the verification result of the document to be verified after inputting the target text information.

7. The multilingual document verification method according to claim 1, characterized in that, The step of obtaining the corresponding term mapping library based on the initial text information includes: Obtain the target region and document type corresponding to the initial text information; A corresponding terminology mapping library is obtained based on the target region and the document type; wherein, the terminology mapping library is a data mapping library established in advance according to the terms corresponding to each document type in each preset region.

8. A multilingual document verification device, characterized in that, The device includes: The segmentation module is used to segment the scanned image of the document to be verified using a preset segmentation model to obtain multiple text region images; The recognition module is used to identify the language type corresponding to each of the text region images; The acquisition module is used to acquire the corresponding document processing method based on the language type of each of the text region images; The processing module is used to process the corresponding text region images based on the document processing method to obtain the initial text information corresponding to each text region image; The mapping module is used to obtain the corresponding term mapping library based on the initial text information, and to map the initial text information using the corresponding term mapping library to obtain the target text information; The verification module is used to verify the target text information to obtain the verification result of the document to be verified.

9. A computer-readable storage medium, characterized in that, The document contains a computer program that, when executed by a processor, causes the processor to perform the steps of the multilingual document verification method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the multilingual document verification method as described in any one of claims 1 to 7.

Images