Control analysis method and device for production document version, equipment, medium and product

Abstract

The invention provides a production document version control analysis method which can be applied to the technical field of artificial intelligence or the field of financial science and technology. The method comprises the following steps: carrying out content category division on pre-acquired historical version commissioning documents to obtain binary files and text files corresponding to the commissioning documents of various versions, carrying out pixel density detection on binary files of adjacent versions to obtain image difference contents, performing multi-modal data comparison on the text files of the adjacent versions to obtain multi-modal difference contents; performing difference analysis on the image difference content and the multi-modal difference content by using a pre-trained intelligent analysis model to obtain image analysis information and multi-modal analysis information; and carrying out information summarization on the image analysis information and the multi-modal analysis information to obtain a version control analysis text. The invention further provides a control analysis device and equipment for the production document version, a medium and a product.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to a control analysis method of a document, and more particularly to a control analysis method and device of a production document version, equipment, medium and product. BACKGROUND

[0002] The system upgrade work of a financial business information system generally needs to be completed within a specific date limited time window, and the task performed during this period is called production. In order to ensure the smooth completion of the production within the production window, the system developers generally make handover of the production content with the operation and maintenance personnel through the production manual, a document medium, before the production day, so that the operation and maintenance personnel can complete the compliance audit while familiarizing themselves with the production content and making early preparation for the production. The production manual can be called a production document, which is usually manually customized and maintained, and for the case of document version update or rollback, the current method is to replace the whole file.

[0003] However, the document version is frequently updated or rolled back, and if the whole file is simply replaced, not only the computer storage resources are wasted, but also the speed of file replacement cannot catch up with the speed of file version update or rollback, thereby leading to low production work efficiency. In the prior art, there are cases of document difference detection with the aid of auxiliary tools, but the current auxiliary tools are limited to identifying the difference of simple paragraphs such as titles and abstracts in the production document, and cannot capture the changes of images, detailed text content, flowcharts and table data in the production document, nor can they provide decision assistance for the staff, resulting in low detection accuracy and poor intelligent level of decision assistance of the current auxiliary tools. SUMMARY

[0004] In view of the above problems, the present application provides a control analysis method, device, equipment, medium and product of a production document version.

[0005] According to a first aspect of the present application, a method for controlling and analyzing a production document version is provided, comprising: dividing a pre-acquired historical version of a production document into content categories to obtain binary files and text files corresponding to each version of the production document, the binary files being used to record image content of the production document, and the text files being used to record text content of the production document; performing pixel density detection on binary files of adjacent versions to obtain image difference content, and performing multi-modal data comparison on text files of adjacent versions to obtain multi-modal difference content, the multi-modal difference content including at least one of text difference content, process difference content, and table data difference content; using a pre-trained intelligent analysis model to perform difference analysis on the image difference content and the multi-modal difference content respectively to obtain image analysis information describing the image difference content and multi-modal analysis information describing the multi-modal difference content; performing information aggregation on the image analysis information and the multi-modal analysis information to obtain version control analysis text describing overall difference between the adjacent versions of the production document and overall difference analysis, and outputting the version control analysis text to a preset user interface.

[0006] According to an embodiment of the present application, pixel density detection is performed on binary files of adjacent versions based on image comparison technology to obtain image difference content, comprising: performing file analysis on binary files of adjacent versions to obtain two initial images; performing preprocessing operation based on pixel density enhancement on the two initial images to obtain two standardized images with pixel density difference; performing three-dimensional feature extraction on the two standardized images respectively to obtain image visual features, image text features, and image structure features representing the binary files of adjacent versions; and performing image difference detection based on the image visual features, the image text features, and the image structure features to obtain the image difference content.

[0007] According to an embodiment of the present application, the image difference content includes image difference regions, image difference expressions, and image difference components; wherein the image difference regions are obtained by calculating visual similarity based on image visual features of adjacent versions and screening regions with visual similarity values lower than a preset similarity threshold in the standardized images; the image difference expressions are obtained by finding a longest common subsequence based on image text features of adjacent versions and retaining text expressions corresponding to sequences other than the longest common subsequence in text expressions of the standardized images; and the image difference components are obtained by component recognition based on image structure features of adjacent versions through a preset tree matching algorithm.

[0008] According to the embodiment of the present application, the multi-modal data comparison is performed on the adjacent version of the text type file to obtain the multi-modal difference content, including: performing the format detection on the adjacent version of the text type file, confirming the file type of the text type file, and obtaining the text type format separator about the file type; wherein the file type includes the lightweight markup language format, the extensible markup language format, and the table data format; in response to receiving the confirmation instruction about the format unification, the separator is removed in the text type file to unify the format of the text type file into the pure text file; the preset text conflict algorithm is used to perform the text difference detection based on the line comparison on the adjacent version of the pure text file to obtain the multi-modal difference content.

[0009] According to the embodiment of the present application, the intelligent analysis model is a large model, and the machine vision model for image difference content and the machine text model for multi-modal difference content are integrated in the large model; wherein the difference rationality analysis is performed on the image difference content by the machine vision model to obtain the image analysis information, which includes: obtaining the image difference area, the image difference expression, and the image difference component of the image difference content; the key point detection is performed on the image difference area to obtain the first interface element of the adjacent version of the binary file in the image difference area, and the first interface element includes at least one of menu, table, and index; the object variant recognition is performed on the area related to the image difference component in the binary file to obtain the second interface element of the adjacent version in the binary file, and the second interface element includes at least one of button, input box, and icon; the expression object is determined according to the image difference expression, and the consistency judgment is respectively performed on the expression object and the first interface element and the second interface element to obtain the consistency coefficient, and the difference rationality expression is performed based on the consistency coefficient to obtain the image analysis information; wherein the consistency coefficient is used to represent the degree of difference rationality, and the higher the consistency coefficient is, the stronger the difference rationality is.

[0010] According to the embodiment of the present application, the difference analysis is performed on the multi-modal difference content by the machine text model, including: determining the file type of the text type file to which the multi-modal difference content belongs; performing the targeted difference analysis on the multi-modal difference content according to the file type to obtain the multi-modal analysis information corresponding to the file type, and the multi-modal analysis information includes at least one of the lightweight language analysis information, the process description text, and the table data rationality information.

[0011] According to an embodiment of the present application, when the file type is a lightweight markup language format, the difference of the multi-modal difference content is analyzed for rationality to obtain lightweight language analysis information, which includes: obtaining semantic embedding distance of two pure text files involved in the multi-modal difference content; performing ambiguity judgment according to the semantic embedding distance to obtain an ambiguity coefficient of the multi-modal difference content, and performing difference rationality expression based on the ambiguity coefficient to obtain the lightweight language analysis information; wherein the ambiguity coefficient is used to represent the degree of difference rationality, and the higher the ambiguity coefficient, the stronger the difference rationality.

[0012] According to an embodiment of the present application, when the file type is an extensible markup language format, the multi-modal difference content is analyzed for difference to obtain flow description information, which includes: obtaining an extensible markup language flowchart of a text file involved in the multi-modal difference content; performing structural conversion on the extensible markup language flowchart to obtain a flow structure containing events and edges; performing feature extraction on the flow graph structure to obtain a flow embedding representation; and generating a flow description text in a specific format according to the flow embedding representation.

[0013] According to an embodiment of the present application, when the file type is a table data format, the multi-modal difference content is analyzed for difference to obtain table data rationality information, which includes: obtaining table data of adjacent versions involved in the multi-modal difference content; obtaining historical data corresponding to a table header of the table data, determining a change trend based on the historical data, and determining a rationality change interval according to the change trend; calculating a deviation coefficient according to a difference value of the table data of the adjacent versions and the rationality change interval, and performing difference rationality expression according to the deviation coefficient to obtain the table data rationality information; wherein the deviation coefficient is used to represent the degree of deviation of the difference value from the rationality change interval, and the smaller the deviation coefficient, the stronger the difference rationality.

[0014] The second aspect of the present application provides a control analysis device for a production document version, comprising: a content division module configured to divide a historical version of a production document into content categories to obtain a binary file and a text file corresponding to each version of the production document, the binary file being configured to record image content of the production document, and the text file being configured to record text content of the production document; a difference acquisition module configured to perform pixel density detection on binary files of adjacent versions to obtain image difference content, and perform multi-modal data comparison on text files of adjacent versions to obtain multi-modal difference content, the multi-modal difference content including at least one of text difference content, process difference content, and table data difference content; a difference analysis module configured to use a pre-trained intelligent analysis model to perform difference analysis on the image difference content and the multi-modal difference content respectively to obtain image analysis information describing the image difference content and multi-modal analysis information describing the multi-modal difference content; and an information output module configured to perform information aggregation on the image analysis information and the multi-modal analysis information to obtain version control analysis text describing overall difference between the adjacent versions of the production document and overall difference analysis, and output the version control analysis text to a preset user interface.

[0015] The third aspect of the present application provides an electronic device, comprising: one or more processors; a memory configured to store one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method.

[0016] The fourth aspect of the present application further provides a computer-readable storage medium having a computer program or instructions stored thereon, wherein the computer program or instructions are executed by a processor to implement the steps of the method.

[0017] The fifth aspect of the present application further provides a computer program product comprising a computer program or instructions, wherein the computer program or instructions are executed by a processor to implement the steps of the method.

[0018] The one or more embodiments have the beneficial effects that the historical version production document is divided into a binary file recording image content and a text file recording text content, and the differences between versions are detected by focusing on images and text, so that not only the comparison between texts but also the comparison between images can be involved, and the comparison between the text file and the multi-modal text file involves the comparison between the multi-modal difference content including text difference content, process difference content and table data difference content, so as to improve the coverage of document version content comparison and the accuracy of document version difference identification; and the multi-modal difference content is further analyzed by the pre-trained intelligent analysis model to obtain version control analysis text, so that the user can intuitively obtain the difference content and analysis description of the difference content in the user interface, thereby improving the intelligent level of version control and the intelligent level of decision-making assistance. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above content and other purposes, features and advantages of the present application will be more apparent from the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0020] Figure 1 An application scenario diagram of a production document version control analysis method, device, equipment, medium and program product according to an embodiment of the present application is schematically shown;

[0021] Figure 2 A flowchart of a production document version control analysis method according to an embodiment of the present application is schematically shown;

[0022] Figure 3 A data flow direction schematic diagram involved in a production document version control analysis method according to an embodiment of the present application is schematically shown;

[0023] Figure 4 A functional component schematic diagram involved in a production document version control analysis method according to an embodiment of the present application is schematically shown;

[0024] Figure 5 A structural block diagram of a production document version control analysis device according to an embodiment of the present application is schematically shown; and

[0025] Figure 6 A block diagram of an electronic device suitable for implementing a production document version control analysis method according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0026] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0027] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0029] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0030] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0031] Existing technologies use auxiliary tools to detect differences between different versions of production documents. However, current auxiliary tools are limited to identifying differences in simple paragraphs such as titles and summaries in production documents. They cannot capture changes in images, detailed text content, flowcharts, and table data in production documents, nor can they provide decision support for staff. As a result, current auxiliary tools not only have low detection accuracy but also poor level of intelligent decision support.

[0032] This application provides a method, apparatus, device, medium, and product for controlling and analyzing production document versions. The method includes: classifying pre-acquired historical versions of production documents into content categories to obtain binary files and text files corresponding to each version; the binary files record image content of the production documents, and the text files record text content; performing pixel density detection on adjacent binary files to obtain image difference content; comparing adjacent text files using multimodal data to obtain multimodal difference content, where multimodal difference content includes at least one of text difference content, process difference content, and table data difference content; performing difference analysis on the image difference content and multimodal difference content using a pre-trained intelligent analysis model to obtain image analysis information describing the image difference content and multimodal analysis information describing the multimodal difference content; summarizing the image analysis information and multimodal analysis information to obtain version control analysis text describing the overall differences and overall difference analysis between adjacent versions of production documents; and outputting the version control analysis text to a preset user interface.

[0033] Figure 1 The diagram illustrates an application scenario of the control and analysis method for production document versions according to an embodiment of this application.

[0034] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105.

[0035] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0036] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0037] Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables. Network 104 serves as the medium for providing communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0038] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0039] It should be noted that the production document version control and analysis method provided in this application embodiment can generally be executed by server 105. Correspondingly, the production document version control and analysis device provided in this application embodiment can generally be located in server 105. The production document version control and analysis method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the production document version control and analysis device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0040] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0041] The following will be based on Figure 1 The described scene, through Figures 2-4 The method for controlling and analyzing production document versions according to embodiments of this application will be described in detail.

[0042] Figure 2 A flowchart illustrating a control analysis method for production document versions according to an embodiment of this application is shown schematically.

[0043] like Figure 2 As shown, the control and analysis method for the production document version in this embodiment includes steps S210 to S250, as detailed below:

[0044] Operation S210 categorizes the pre-acquired historical production documents into content categories, obtaining binary files and text files corresponding to each version of the production document. The binary files record the image content of the production document, and the text files record the text content of the production document. Operation S220 performs pixel density detection on the binary files of adjacent versions to obtain image difference content, and performs multimodal data comparison on the text files of adjacent versions to obtain multimodal difference content. The multimodal difference content includes at least one of text difference content, process difference content, and table data difference content. Operation S230 uses a pre-trained intelligent analysis model to perform difference analysis on the image difference content and the multimodal difference content, respectively, to obtain image analysis information describing the image difference content and multimodal analysis information describing the multimodal difference content. Operation S240 summarizes the image analysis information and the multimodal analysis information to obtain version control analysis text describing the overall differences and overall difference analysis between adjacent versions of the production document, and outputs the version control analysis text to a preset user interface.

[0045] As an example, the binary file can be an image file, such as PNG (Portable Network Graphics), JPEG (JPG, Joint Photographic Experts Group), etc., which are used in production documentation to record explanatory images; the text file can be any file format capable of recording text, such as CSV (Comma-Separated Values), also known as tabular data format, which is a text file used to store tabular data, typically used in production documentation to record parameter tables; or lightweight markup languages, also known as lightweight markup language formats, which are often used in production documentation to record natural language text; XML (eXtensible Markup Language)... The production documentation typically records process diagrams in the Natural Language (ML) format. After obtaining the aforementioned binary and text files, data comparison or detection is performed on them to identify image differences and multimodal differences. Then, difference analysis is performed on these differences to obtain version control analysis text describing the overall differences between adjacent versions of the production documentation and the overall difference analysis. The entire detection and analysis process involves the natural language text used to explain the implementation of production steps, the process diagrams used to visually demonstrate the process logic, the explanatory images used to provide practical operation references, and the parameter tables used to collect system maintenance parameters, thus covering the entire content of a production documentation.

[0046] Therefore, the production document version control and analysis method of this embodiment not only involves the comparison between texts but also the comparison between images. Moreover, for text files, it involves the comparison of text files of various formats to obtain multimodal difference content, including text difference content, process difference content, and table data difference content. This improves the coverage of document version content comparison and thus improves the accuracy of document version difference identification. Furthermore, based on the multimodal difference content, further intelligent analysis is performed to obtain analysis information, allowing users to intuitively obtain the difference content and analysis description of the difference changes on the user interface, thereby improving the intelligence level of version control.

[0047] Figure 3 The diagram illustrates the data flow in the control analysis method for production document versions according to an embodiment of this application.

[0048] like Figure 3 As shown, in this embodiment, pixel density detection is performed on adjacent versions of binary files based on image comparison technology to obtain image difference content. This includes: parsing the adjacent versions of binary files to obtain two initial images; performing a pixel density enhancement-based preprocessing operation on the two initial images to obtain two standardized images with pixel density differences; performing three-dimensional feature extraction on the two standardized images to obtain image visual features, image text features, and image structural features that characterize the adjacent versions of binary files; and performing image difference detection based on the image visual features, image text features, and image structural features to obtain image difference content.

[0049] As an example, pixel density detection is performed on adjacent versions of binary files to obtain image differences, and multimodal data comparison is performed on adjacent versions of text files to obtain multimodal differences. Both are executed by a preset version control engine, which is the core processing component. To implement fine-grained version control of document content, natural language text can be version controlled based on lines, flowcharts can be version controlled based on nodes, explanatory images can be version controlled based on local pixels, and parameter tables can be version controlled based on parameter values. In principle, text files should be used as much as possible rather than binary files to facilitate text layout, flexible adjustment of chart nodes, and focused parameter changes. Therefore, natural language text can use lightweight markup language format files as information carriers, flowcharts are required to use XML format files as information carriers, parameter tables are required to use CSV format files as information carriers, and explanatory images need to be stored as bitmaps, so PNG, JPG, and JPEG files are all supported.

[0050] In this specific example, for binary files, after obtaining two versions of the binary files, the files are parsed to obtain two initial images. These images are then preprocessed, including cropping and noise reduction. To facilitate visual differentiation between the two initial images, a pixel density enhancement preprocessing operation is performed. This operation primarily involves deepening the pixels in densely populated regions, thus making it easier to distinguish the differences in pixel density between the two images and obtaining two standardized images. Subsequently, image processing and feature extraction are performed using machine vision and OCR technologies. Then, version difference detection, annotation, and analysis are performed in this component to obtain the image difference content.

[0051] Based on this, the two images are first preprocessed in terms of pixel density to highlight the differences between them. Then, feature extraction is performed to detect the differences in the images, thereby improving the detection accuracy of image difference detection. Furthermore, the control analysis of production document versions is not limited to the text title stage, but can also perform difference detection and analysis on the images in the production document, thereby improving the feature focus and coverage of version control and thus improving the difference detection accuracy in the version control process.

[0052] In this embodiment, the image difference content includes image difference regions, image difference descriptions, and image difference components. The image difference regions are obtained by calculating visual similarity based on the visual features of adjacent versions of the image, and by selecting regions in the standardized image whose visual similarity values ​​are lower than a preset similarity threshold. The image difference descriptions are obtained by finding the longest common subsequence based on the text features of adjacent versions of the image, and by retaining the text descriptions corresponding to sequences other than the longest common subsequence in the text descriptions of the standardized image. The image difference components are obtained by identifying components based on the structural features of adjacent versions of the image using a preset tree matching algorithm.

[0053] As an example, for image visual features, the similarity of visual features is calculated through a contrastive learning model, focusing on the difference regions within the image; for image text features, text conflict algorithms, such as a pre-defined text diff algorithm, are used to compare OCR results, i.e., comparing the differences in text within the image text features. In addition to the text itself, the coordinates of the text regions also need to be compared so as to map back to the image position; for image structural features, tree matching algorithms are used to compare the edit distance of the layout tree, thereby identifying added, deleted, or moved UI components, and then locating the changed areas. Through an attention-based change density detection network, it is possible to automatically focus on high-probability change areas. For example, based on the above high-density preprocessing, it is possible to discover densely interactive areas of UI components. Finally, the network generates a difference heatmap, which can be used as the image difference content.

[0054] Therefore, it can not only focus on differences in the surface area of ​​an image, but also on image text with differences, as well as components in the image, further improving the accuracy and coverage of image difference detection.

[0055] Figure 4 This illustration schematically shows a diagram of the functional components involved in the control and analysis method for production document versions according to an embodiment of this application.

[0056] exist Figure 4 In the illustrated embodiment, after the version control engine obtains the image difference content, it can directly transmit the image difference content to the user interface for the user to view and use. To facilitate the user's viewing of the image difference content, before transmitting it to the user interface, after obtaining the image difference content (including the similarity of various types of regions), these difference changes need to be aggregated and classified. Otherwise, this information would be too cumbersome for users of the document version control system. First, aggregation is performed. Since the difference changes of different adjacent image sub-regions may be similar, it is necessary to cluster adjacent sub-regions with similar difference changes and merge their difference changes to avoid fragmented output. Then, classification is performed. Using a machine learning-based classifier, such as the random forest method, the difference changes are divided into the following three categories: 1) Content changes, such as changes in text content or changes in highlighted annotations. 2) Structural changes: such as the addition or removal of elements in the operation interface. 3) Style changes, such as adjustments to color and font styles. Finally, by using pre-set expert rules, the image difference content is marked according to the changes in different types of classifications, including quantity and type, and a difference change report with a structured representation of the image difference content is generated. The marked image difference content and the difference change report are then transmitted to the user interface.

[0057] In this embodiment, multimodal data comparison is performed on adjacent versions of text files to obtain multimodal differences. This includes: performing format detection on adjacent versions of text files to confirm the file type of the text files and obtaining text format delimiters related to the file type; wherein, the file type includes Lightweight Markup Language (LTL), Extensible Markup Language (EXPLAIN), and tabular data format; in response to receiving a confirmation instruction regarding format consistency, delimiters are removed from the text files to unify the format of the text files into plain text files; and a preset text conflict algorithm is used to perform line-by-line comparison-based text difference detection on adjacent versions of plain text files to obtain multimodal differences.

[0058] As an example, the operation of comparing adjacent versions of text files in a multimodal manner to obtain multimodal differences is also performed by the version control engine. For text files, due to different formats such as Lightweight Markup Language (Lightweight Markup Language), XML (eXtensible Markup Language), and CSV (Comma-Separated Values), also known as tabular data format, the inconsistent formats require unification into plain text files to facilitate the detection of text changes. However, the version control engine cannot directly perform format unification conversion; it needs to send a format conversion request to the user interface first. When the user agrees to the format conversion, the version control engine receives a confirmation instruction. In response to receiving the confirmation instruction regarding format unification, it removes delimiters from the text files to unify the format of the text files into plain text files. Then, it performs text difference detection to obtain multimodal difference content, which can then be transmitted to the user interface for the user to view intuitive multimodal difference information.

[0059] Therefore, upon receiving the confirmation instruction for format consistency, the above format is uniformly processed into a plain text file to facilitate the acquisition of differences between texts and improve the accuracy of text difference detection.

[0060] In this embodiment, the intelligent analysis model is a large model that integrates a machine vision model for image difference content and a machine text model for multimodal difference content. Specifically, the machine vision model performs a reasonableness analysis on the image difference content to obtain image analysis information, including: acquiring the image difference region, image difference description, and image difference components; performing key point detection on the image difference region to obtain first interface elements that change in adjacent versions of the binary file within the image difference region, the first interface element including at least one of a menu, table, and index; performing object variant recognition on the regions involved in the image difference components within the binary file to obtain second interface elements that occur in adjacent versions of the binary file, the second interface element including at least one of a button, input box, and icon; determining the description object based on the image difference description, and performing consistency judgments on the description object with the first and second interface elements to obtain a consistency coefficient; and performing a reasonableness description of the difference based on the consistency coefficient to obtain image analysis information. The consistency coefficient characterizes the degree of reasonableness of the difference; the higher the consistency coefficient, the stronger the reasonableness of the difference.

[0061] As an example, this large model refers to an artificial intelligence large model, a type of artificial intelligence model with a large number of parameters constructed by artificial neural networks. Such large AI models typically include large language models, large vision models, multimodal large models, and basic science large models. In this embodiment, the large models include machine vision models and machine text models with different functions. The machine vision model is the large vision model. This large vision model can integrate an OCR ensemble model with image processing capabilities, a language model with text processing capabilities, and an object detection model for specific object detection. The OCR ensemble model extracts all text content and its image coordinates from the image, including button labels and titles, thereby performing keypoint detection on image difference regions and obtaining the first change in the image difference regions between adjacent versions of the binary file. Interface elements; through object detection models, such as identifying UI components (e.g., buttons, input boxes, icons), and constructing layout trees or grid representations, object variant identification is performed on the regions involved by image difference components in the binary file to obtain the second interface elements that occur in adjacent versions of the binary file; through language models, such as converting text into embedded representations, the representation object is determined based on the image difference description, and consistency judgments are made between the representation object and the first and second interface elements to obtain consistency coefficients. Based on the consistency coefficients, the reasonableness of the difference is described to obtain image analysis information; the machine text model in this embodiment is essentially a large language model. The operation process and functional modules of this large language model will be described in detail in the subsequent description of the operation steps of performing difference analysis on multimodal difference content through the machine text model.

[0062] Thus, the large model is divided into two parts: a machine vision model for image difference analysis and a machine text model for text difference analysis. Based on this, it can make a reasonable assessment of the image differences between the two versions of the production text, thereby obtaining a reasonable description of the differences. This makes it easier for users to intuitively understand the difference descriptions between the two versions of the production document and assist them in making decisions about the production document.

[0063] In this embodiment, the machine text model is used to perform differential analysis on multimodal differential content, including: determining the file type of the text file to which the multimodal differential content belongs; performing targeted differential analysis on the multimodal differential content according to the file type to obtain multimodal analysis information corresponding to the file type, wherein the multimodal analysis information includes at least one of lightweight language analysis information, process description text, and table data rationality information.

[0064] As an example, for text files of different file types, targeted analysis is performed on the multimodal differences generated, such as difference rationality analysis, or directly obtaining the description of the difference content itself. For example, CSV format, i.e., a tabular data format, is a text file used to store tabular data, usually recording parameter tables in production documents. Therefore, the difference content generated by text files of this format is the tabular data difference content, and the analysis of this tabular data difference content yields the tabular data rationality information. Lightweight markup language format is often used to record natural language text in production documents. Therefore, the difference content corresponding to this format is the text difference content, and the analysis of this text difference content yields lightweight language analysis information. XML format (Extensible Markup Language) format is usually used to record process diagrams in production documents. The difference content generated by the corresponding text files is the process difference content, and the analysis of this process difference content yields the process description text.

[0065] In other words, the three types of multimodal analysis information—lightweight language analysis information, process description text, and table data rationality information—correspond to the three types of multimodal difference content: text difference content, process difference content, and table data difference content, respectively. This establishes a targeting between the types of multimodal analysis information and the types of text files, thereby enabling different targeted difference analyses to be performed based on the file type. This improves the coverage and effectiveness of multimodal analysis information, and ultimately enhances the intelligent level of version control's auxiliary decision-making.

[0066] In this embodiment, when the file type is a lightweight markup language format, a reasonableness analysis of the differences in multimodal difference content is performed to obtain lightweight language analysis information. This includes: obtaining the semantic embedding distance between the two plain text files involved in the multimodal difference content; determining ambiguity based on the semantic embedding distance to obtain the ambiguity coefficient of the multimodal difference content; and expressing the reasonableness of the differences based on the ambiguity coefficient to obtain lightweight language analysis information. The ambiguity coefficient is used to characterize the degree of reasonableness of the differences; the higher the ambiguity coefficient, the stronger the reasonableness of the differences.

[0067] As an example, a machine text model is essentially a large language model, integrating functional modules for lightweight markup language formats, modules for differential analysis of lightweight markup language formats, and modules for differential analysis of tabular data formats. This specific example demonstrates the process of differential analysis for multimodal differential content in text. During the pre-training of the lightweight markup language format functional modules within the machine text model, a perturbation model can be used to automatically add, replace, or delete local text content based on prompt words using random masks, generating slight variations of the text (e.g., generating "restart service" as "restart database service"). Then, another pre-trained language architecture is used as the detection model to compare the semantic embedding distance between the original and variant texts. If the distance exceeds the threshold, it indicates that the original text is sensitive to minor changes and may be ambiguous. In addition, to enhance scenario adaptability, common terms involved in the production support scenarios of the financial business information system will be maintained in the dictionaries of the interference model and the detection model, and will be given higher weights during the generation process. During the adversarial fine-tuning learning process, if the detection threshold exceeds the expected setting (too high or too low), human intervention will be used for positive feedback correction. Finally, the detection model trained is used as a functional module for lightweight markup language formats in the machine text model for text ambiguity detection.

[0068] Therefore, the higher the ambiguity coefficient, the greater the semantic difference between the two documents, and the more distinctive the version update. This makes it easier for users to understand the degree of textual differences, thereby improving the level of intelligence in version control decision support.

[0069] In this embodiment, when the file type is Extensible Markup Language (XML) format, differential analysis is performed on the multimodal differential content to obtain process description information, including: obtaining the XML flowchart of the text-type files involved in the multimodal differential content; performing structural transformation on the XML flowchart to obtain a process structure containing events and edges; extracting features from the flowchart structure to obtain a process embedding representation; and generating process description text in a specific format based on the process embedding representation.

[0070] As an example, this describes the process of obtaining process description information by performing differential analysis on multimodal differential content, namely process differential content. In the machine text model, the module for differential analysis of lightweight markup language format can first parse the XML flowchart into a semantic graph structure based on a graph attention network, containing nodes (such as conditions and events) and edges (such as sequence and loop). During the training phase, this module for differential analysis of lightweight markup language format learns the vector representations of nodes and edges, that is, the graph embedding representation that encodes the graph structure and contextual information. Then, based on the process specification corpus, the pre-trained model is fine-tuned to learn and become familiar with the language style and idiomatic expressions of process description (such as "when...", "if...then..."), thereby understanding common patterns of process logic, such as sequence, branching, concurrency, looping, and exceptions. When generating summaries that match the flowchart, the graph embedding representation is taken as input. Its internal attention mechanism focuses on the part of the graph embedding most relevant to the currently generated word (specific nodes, edges, or relationships), ensuring that each step of the generated description is anchored to a specific part of the semantic graph structure. Furthermore, to ensure that the nodes or logical patterns mentioned in the generated summary are consistent with the actual flowchart, constraint decoding is applied during the generation process. The final trained text summarization model will be used to convert flowcharts into flowchart text summaries.

[0071] Based on this, flowcharts in XML format can be converted into descriptive text, allowing users to intuitively understand the differences between flowcharts in two production documents, thereby improving the intelligence level of version control's auxiliary decision-making.

[0072] In this embodiment, when the file type is tabular data format, a difference analysis is performed on the multimodal difference content to obtain the reasonableness information of the tabular data. This includes: obtaining the tabular data of adjacent versions involved in the multimodal difference content; obtaining the historical data corresponding to the table header of the tabular data, determining the change trend based on the historical data, and determining the reasonable change range based on the change trend; calculating the deviation coefficient based on the difference between adjacent versions of the tabular data and the reasonable change range, and expressing the reasonableness of the difference based on the deviation coefficient to obtain the reasonableness information of the tabular data. The deviation coefficient is used to characterize the degree to which the difference deviates from the reasonable change range. The smaller the deviation coefficient, the stronger the reasonableness of the difference.

[0073] As an example, this describes the process of performing difference analysis on multimodal differences in table data. The machine text model's module for difference analysis of table data formats can treat all parameters of the table data differences as a spatiotemporal data stream, viewing the parameters from both temporal and spatial perspectives. Firstly, there's the temporal context: the full parameter table of the current production document serves as a snapshot, and the parameter modifications mentioned in the document represent future changes. Meanwhile, the parameters of each previous version of the business system serve as past references. The time series model within the machine text model's module for difference analysis of table data formats analyzes the changing trends of parameter values ​​and detects whether future parameter changes "drift" out of a reasonable range. Secondly, there's the spatial context: since all parameter values ​​have their modification basis, such as specifications, recommendations, or business requirements, the pre-trained time series model treats these basis sources as context, using them as training data for fine-tuning and learning the ability to judge parameter rationality. The resulting spatiotemporal context-aware model is then used as part of the machine text model's module for difference analysis of table data formats to determine whether the parameters to be modified are reasonable and effective.

[0074] Therefore, it is possible to conduct a rationality analysis of changes in the number of tables. Furthermore, since the table data in production documents generally does not change significantly, the smaller the deviation, the stronger the rationality of the difference. This allows users to intuitively understand the changes in table data, promptly detect data errors, and improve the intelligent level of version control's auxiliary decision-making.

[0075] As described above, the production document version control analysis method provided in this embodiment divides historical production documents into binary files containing image content and text files containing text content. It focuses on detecting differences between versions of both images and text, thus encompassing not only text-to-text comparisons but also image-to-image comparisons. Furthermore, the text files involve comparisons of multimodal formats, yielding multimodal difference content including textual differences, process differences, and table data differences. This improves the coverage of document version content comparison and consequently enhances the accuracy of document version difference identification. Moreover, a pre-trained intelligent analysis model further intelligently analyzes the multimodal difference content to obtain version control analysis text. This allows users to intuitively obtain information about the changes in differences and their analysis descriptions on the user interface, thereby improving the intelligence level of version control and decision support.

[0076] Based on the above-described method for controlling and analyzing production document versions, this application also provides a device for controlling and analyzing production document versions. The following will be combined with... Figure 5 The device is described in detail.

[0077] Figure 5A schematic block diagram of a control and analysis apparatus for a production documentation version according to an embodiment of this application is shown.

[0078] like Figure 5 As shown, the production document version control and analysis device 500 of this embodiment includes a content division module 510, a difference acquisition module 520, a difference analysis module 530, and an information output module 540.

[0079] The content segmentation module 510 can perform operation S210, which is used to segment the content categories of the pre-acquired historical version production documents to obtain binary files and text files corresponding to each version of the production documents. The binary files are used to record the image content of the production documents, and the text files are used to record the text content of the production documents.

[0080] The difference acquisition module 520 can perform operation S220 to perform pixel density detection on adjacent versions of binary files to obtain image difference content, and to perform multimodal data comparison on adjacent versions of text files to obtain multimodal difference content. The multimodal difference content includes at least one of text difference content, process difference content, and table data difference content.

[0081] The difference analysis module 530 can perform operation S230 to perform difference analysis on image difference content and multimodal difference content using a pre-trained intelligent analysis model, so as to obtain image analysis information describing image difference content and multimodal analysis information describing multimodal difference content.

[0082] The information output module 540 can perform operation S240 to summarize image analysis information and multimodal analysis information to obtain version control analysis text that describes the overall differences between adjacent production documents and the overall difference analysis, and output the version control analysis text to a preset user interface.

[0083] In this embodiment, the difference acquisition module 520 includes an image detection unit and a text detection unit.

[0084] This image detection unit is used to perform pixel density detection on adjacent versions of binary files based on image comparison technology to obtain image difference content. The process includes: parsing the adjacent versions of binary files to obtain two initial images; performing pixel density enhancement preprocessing on the two initial images to obtain two standardized images with pixel density differences; extracting three-dimensional features from the two standardized images to obtain image visual features, image text features, and image structural features representing the adjacent versions of binary files; and performing image difference detection based on the image visual features, image text features, and image structural features to obtain image difference content. This image difference content includes image difference regions, image difference descriptions, and image difference components. Specifically, image difference regions are obtained by calculating visual similarity based on the visual features of adjacent versions of the images, and regions with visual similarity values ​​below a preset similarity threshold are selected from the standardized images. Image difference descriptions are obtained by finding the longest common subsequence based on the image text features of adjacent versions, and retaining the text descriptions corresponding to sequences other than the longest common subsequence in the text descriptions of the standardized images. Image difference components are obtained by component identification based on the image structural features of adjacent versions using a preset tree matching algorithm.

[0085] The text detection unit performs multimodal data comparison on adjacent versions of text files to obtain multimodal differences. This includes: performing format detection on adjacent versions of text files to confirm the file type of the text files and obtaining text format delimiters related to the file type; wherein the file type includes Lightweight Markup Language (LTL), Extensible Markup Language (EXPLAIN), and tabular data format; in response to receiving a confirmation instruction regarding format consistency, removing delimiters from the text files to unify the format of the text files into plain text files; and using a preset text conflict algorithm to perform line-by-line comparison-based text difference detection on adjacent versions of plain text files to obtain multimodal differences.

[0086] In this embodiment, the intelligent analysis model in the difference analysis module 530 is a large model, which integrates a machine vision model for image difference content and a machine text model for multimodal difference content. Specifically, the machine vision model performs a reasonableness analysis on the image difference content to obtain image analysis information, including: acquiring the image difference region, image difference description, and image difference components; performing key point detection on the image difference region to obtain the first interface element that changes in the image difference region in adjacent versions of the binary file, the first interface element including at least one of a menu, table, and index; performing object variant recognition on the region involved in the image difference component in the binary file to obtain the second interface element that occurs in the binary file in adjacent versions, the second interface element including at least one of a button, input box, and icon; determining the description object based on the image difference description, and performing consistency judgment between the description object and the first and second interface elements to obtain a consistency coefficient; and performing a reasonableness description of the difference based on the consistency coefficient to obtain image analysis information. The consistency coefficient characterizes the degree of reasonableness of the difference; the higher the consistency coefficient, the stronger the reasonableness of the difference.

[0087] The machine text model is used to perform differential analysis on multimodal differential content, including: determining the file type of the text file to which the multimodal differential content belongs; performing targeted differential analysis on the multimodal differential content according to the file type to obtain multimodal analysis information corresponding to the file type, which includes at least one of lightweight language analysis information, process description text, and table data rationality information; wherein, when the file type is in lightweight markup language format, a rationality analysis of the differential content is performed to obtain lightweight language analysis information, including: obtaining the semantic embedding distance between the two plain text files involved in the multimodal differential content; performing ambiguity judgment based on the semantic embedding distance to obtain the ambiguity coefficient of the multimodal differential content, and expressing the rationality of the difference based on the ambiguity coefficient to obtain lightweight language analysis information; wherein, the ambiguity coefficient is used to characterize the degree of rationality of the difference, and the higher the ambiguity coefficient, the stronger the rationality of the difference; when the file type is in extensible markup language format, the multimodal differential content is analyzed... The process involves difference analysis to obtain process description information, including: acquiring the Extensible Markup Language (XML) flowcharts of text files involved in the multimodal difference content; performing structural transformation on the XML flowcharts to obtain a process structure containing events and edges; extracting features from the flowchart structure to obtain a process embedding representation; generating process description text in a specific format based on the process embedding representation; and, in the case of tabular data, performing difference analysis on the multimodal difference content to obtain tabular data rationality information, including: acquiring adjacent versions of tabular data involved in the multimodal difference content; acquiring historical data corresponding to the table headers, determining the trend of change based on the historical data, and determining the rationality range of change based on the trend of change; calculating the deviation coefficient based on the difference between adjacent versions of tabular data and the rationality range of change, and expressing the rationality of the difference based on the deviation coefficient to obtain the rationality information of the tabular data; wherein, the deviation coefficient is used to characterize the degree to which the difference deviates from the rationality range of change, the smaller the deviation coefficient, the stronger the rationality of the difference.

[0088] Furthermore, according to embodiments of this application, any multiple modules among the content partitioning module 510, difference acquisition module 520, difference analysis module 530, and information output module 540 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the content partitioning module 510, difference acquisition module 520, difference analysis module 530, and information output module 540 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any appropriate combination of any of these three implementation methods. Alternatively, at least one of the content segmentation module 510, difference acquisition module 520, difference analysis module 530, and information output module 540 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0089] It should be noted that the implementation methods, technical problems solved, functions achieved, and technical effects of each module in the device embodiment are the same as or similar to the implementation methods, technical problems solved, functions achieved, and technical effects of each corresponding step in the method embodiment, and will not be repeated here.

[0090] Figure 6 A block diagram schematically illustrates an electronic device suitable for implementing a control analysis method for production document versions according to an embodiment of this application.

[0091] like Figure 6 As shown, an electronic device 600 according to an embodiment of this application includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage portion 608 into a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0092] RAM 603 stores various programs and data required for the operation of electronic device 600. Processor 601, ROM 602, and RAM 603 are interconnected via bus 604. Processor 601 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 602 and / or RAM 603. It should be noted that the programs may also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0093] According to embodiments of this application, the electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to a bus 604. The electronic device 600 may also include one or more of the following components connected to the input / output (I / O) interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 610 as needed so that computer programs read from it can be installed into the storage section 608 as needed.

[0094] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0095] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 602 and / or RAM 603 and / or one or more memories other than ROM 602 and RAM 603 described above.

[0096] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the control and analysis method for the production documentation version provided in the embodiments of this application.

[0097] When the computer program is executed by the processor 601, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0098] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via the communication section 609, and / or installed from the removable medium 611. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0099] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0100] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0101] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0102] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A method for controlling and analyzing production document versions, characterized in that, The method includes: The previously acquired historical production documents are categorized by content type to obtain binary files and text files corresponding to each version of the production documents. The binary files are used to record the image content of the production documents, and the text files are used to record the text content of the production documents. Pixel density detection is performed on adjacent versions of binary files to obtain image difference content. Multimodal data comparison is performed on adjacent versions of text files to obtain multimodal difference content. The multimodal difference content includes at least one of text difference content, process difference content, and table data difference content. The image difference content and the multimodal difference content are analyzed by using a pre-trained intelligent analysis model to obtain image analysis information describing the image difference content and multimodal analysis information describing the multimodal difference content. The image analysis information and the multimodal analysis information are summarized to obtain version control analysis text that describes the overall differences between adjacent production documents and the overall difference analysis. The version control analysis text is then output to a preset user interface.

2. The method for controlling and analyzing production document versions according to claim 1, characterized in that, Image comparison techniques are used to detect pixel density differences between adjacent versions of binary files to obtain image differences, including: File parsing is performed on adjacent versions of binary files to obtain two initial images; The two initial images are preprocessed based on pixel density enhancement to obtain two standardized images with pixel density differences. Three-dimensional feature extraction is performed on the two standardized images respectively to obtain the image visual features, image text features, and image structural features that characterize the adjacent versions of the binary file; Image difference detection is performed based on the image visual features, the image text features, and the image structural features to obtain the image difference content.

3. The method for controlling and analyzing production document versions according to claim 2, characterized in that, The image difference content includes image difference regions, image difference descriptions, and image difference components; wherein, The image difference region is obtained by calculating the visual similarity based on the visual features of adjacent versions of the image, and then filtering out regions in the standardized image whose visual similarity value is lower than a preset similarity threshold. The image difference representation is obtained by finding the longest common subsequence based on the image text features of adjacent versions, and retaining the text representations corresponding to sequences other than the longest common subsequence in the text representation of the standardized image; The image difference component is obtained by identifying components based on the image structural features of adjacent versions using a preset tree matching algorithm.

4. The method for controlling and analyzing production document versions according to claim 3, characterized in that, Multimodal data comparison is performed on adjacent versions of text files to obtain multimodal differences, including: The adjacent versions of text files are subjected to format detection to confirm the file type of the text files and obtain the text format delimiter for the file type; wherein, the file type includes Lightweight Markup Language format, Extensible Markup Language format, and tabular data format; In response to receiving a confirmation instruction regarding format consistency, the delimiter is removed from the text file to format the text file into a plain text file; A pre-defined text conflict algorithm is used to perform line-by-line comparison-based text difference detection on adjacent versions of plain text files to obtain multimodal differences.

5. The method for controlling and analyzing production document versions according to claim 4, characterized in that, The intelligent analysis model is a large-scale model that integrates a machine vision model for the image difference content and a machine text model for the multimodal difference content. The machine vision model performs a rationality analysis on the image difference content to obtain image analysis information, including: The image difference regions, image difference descriptions, and image difference components of the image difference content are obtained. Key point detection is performed on the image difference region to obtain a first interface element that changes in the image difference region in adjacent versions of the binary file, the first interface element including at least one of menu, table, and index; object variant recognition is performed on the region involved by the image difference component in the binary file to obtain a second interface element that changes in the binary file in adjacent versions, the second interface element including at least one of button, input box, and icon. The object of the description is determined based on the image difference description, and the consistency of the object of the description with the first interface element and the second interface element is judged to obtain the consistency coefficient. The reasonableness of the difference is described based on the consistency coefficient to obtain image analysis information. The consistency coefficient is used to characterize the degree of reasonableness of the difference. The higher the consistency coefficient, the stronger the reasonableness of the difference.

6. The method for controlling and analyzing production document versions according to claim 5, characterized in that, The machine text model is used to perform differential analysis on the multimodal differential content, including: Determine the file type of the text file to which the multimodal difference content belongs; Based on the file type, a targeted difference analysis is performed on the multimodal difference content to obtain multimodal analysis information corresponding to the file type. The multimodal analysis information includes at least one of the following: lightweight language analysis information, process description text, and table data rationality information.

7. The method for controlling and analyzing production document versions according to claim 6, characterized in that, When the file type is a Lightweight Markup Language (LTL) format, a reasonableness analysis of the differences in the multimodal difference content is performed to obtain lightweight language analysis information, including: Obtain the semantic embedding distance between the two plain text files involved in the multimodal difference content; Ambiguity is determined based on the semantic embedding distance to obtain the ambiguity coefficient of the multimodal difference content, and the reasonableness of the difference is expressed based on the ambiguity coefficient to obtain lightweight language analysis information; wherein, the ambiguity coefficient is used to characterize the degree of reasonableness of the difference, and the higher the ambiguity coefficient, the stronger the reasonableness of the difference.

8. The method for controlling and analyzing production document versions according to claim 6, characterized in that, When the file type is Extensible Markup Language (XML) format, a difference analysis is performed on the multimodal difference content to obtain process description information, including: Obtain an Extensible Markup Language flowchart for the text files involved in the multimodal difference content; The Extensible Markup Language flowchart is structurally transformed to obtain a flowchart structure containing events and edges; Feature extraction is performed on the flowchart structure to obtain the flowchart embedding representation; The process embedding is used to generate a process description text in a specific format.

9. The method for controlling and analyzing production document versions according to claim 6, characterized in that, When the file type is a tabular data format, a difference analysis is performed on the multimodal difference content to obtain information on the reasonableness of the tabular data, including: Obtain the table data of adjacent versions involved in the multimodal difference content; Obtain the historical data corresponding to the header of the table data, determine the trend of change based on the historical data, and determine the reasonable range of change based on the trend of change; The deviation coefficient is calculated based on the difference between the adjacent versions of the table data and the reasonableness change range, and the reasonableness of the difference is expressed based on the deviation coefficient to obtain the reasonableness information of the table data; wherein, the deviation coefficient is used to characterize the degree to which the difference deviates from the reasonableness change range, and the smaller the deviation coefficient, the stronger the reasonableness of the difference.

10. A control and analysis device for production document versions, characterized in that, The device includes: The content segmentation module is used to segment the content of the pre-acquired historical version production documents into categories, and obtain binary files and text files corresponding to each version of the production documents. The binary files are used to record the image content of the production documents, and the text files are used to record the text content of the production documents. The difference acquisition module is used to perform pixel density detection on adjacent versions of binary files to obtain image difference content, and to perform multimodal data comparison on adjacent versions of text files to obtain multimodal difference content. The multimodal difference content includes at least one of text difference content, process difference content, and table data difference content. The difference analysis module is used to perform difference analysis on the image difference content and the multimodal difference content using a pre-trained intelligent analysis model, to obtain image analysis information describing the image difference content and multimodal analysis information describing the multimodal difference content. The information output module is used to summarize the image analysis information and the multimodal analysis information to obtain version control analysis text that describes the overall differences between adjacent production documents and the overall difference analysis, and outputs the version control analysis text to a preset user interface.

11. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 9.

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

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

Images