Engineering design document compliance inspection method, electronic device and storage medium

Abstract

The present invention discloses a method, electronic device and storage medium for checking the compliance of engineering design documents, which are based on OCR and a large language model. The method includes the following steps: text recognition and extraction of standard documents; text error checking and manual correction of standard texts; structured processing of the corrected standard texts; vectorization processing of core text content and construction of a standard query system; preset standard set ranges corresponding to each chapter of the design document; vectorization and similarity query of user-selected text in a text editor; and providing corresponding standard content to the user. The method can improve the efficiency and accuracy of standard references in documents, facilitate dynamic update management of standard content, and ensure the correctness of referenced standards.

Description

Technical Field

[0001] The present invention belongs to the field of computer vision technology, and in particular relates to a method for checking the compliance of engineering design documents, an electronic device, and a storage medium. Background Art

[0002] In the prior art, in professional design fields such as engineering, the correctness check of normative references in documents mainly relies on manual review methods and keyword-based document search tools.

[0003] The manual review method refers to when designers write design documents, they refer to paper / electronic specifications based on experience or manually check them. This manual review method has the following problems: 1. Low efficiency: As the specification system becomes increasingly large and complex, the number of specifications referenced in design documents is large. Manually checking them one by one is time-consuming and labor-intensive, and the efficiency is extremely low. Especially when the project cycle is tight, this efficiency problem is particularly prominent. 2. Limitations in accuracy: Manual review is affected by subjective factors, fatigue, and differences in understanding of the latest specifications, which may lead to omissions or errors, and it is difficult to update synchronously to reflect the latest version of national standards or industry specifications, thereby affecting the accuracy of the referenced specifications. This may lead to compliance risks in design results. 3. Lack of uniformity and standardization: Different reviewers may have different levels of familiarity with the specifications, resulting in a lack of uniformity in the use of specification provisions, which is not conducive to standardized management of design quality.

[0004] Keyword-based document search tools, such as the built-in search function of PDF readers, have the following shortcomings: this type of search function can usually only perform simple string matching, and cannot structurally understand and manage the content of the specification, cannot understand semantics, cannot identify the structured information of the specification such as entries, figures, tables, etc., cannot understand the true meaning of the text, and cannot determine the current validity of the referenced specification, resulting in the inability to perform accurate semantic matching and content association; each step of the review relies on manual judgment and experience, resulting in the accuracy of the inspection being limited by individual ability, making it easy to make mistakes and lacking unified standards; when cross-references to multiple specifications are involved, it is necessary to switch repeatedly between different documents, and the efficiency is still low. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, electronic device and storage medium for compliance checking of engineering design documents, so as to solve the technical shortcomings of manual review of standard references in documents in the prior art, such as low efficiency, low accuracy, inability to dynamically and effectively manage and lack of uniformity, so as to improve the efficiency and accuracy of standard references in documents, facilitate the dynamic update management of standard content, and ensure the correctness of referenced standards.

[0006] In a first aspect, the present invention provides a method for checking compliance of an engineering design document, comprising the following steps:

[0007] Text recognition and extraction of specification documents: Use optical character recognition technology to recognize and extract text from engineering specification documents to be consulted, forming editable and searchable specification texts;

[0008] Error checking and manual correction of standard texts: A large language model is introduced to check and prompt errors in the standard texts, and the errors are manually corrected;

[0009] Structural processing of the revised standard text: identifying and extracting specific information units in the standard text to obtain a structured standard data set, wherein each specific information unit in the standard data set is accompanied by corresponding physical location information and logical location information;

[0010] Vectorization of core text content and establishment of a standardized query system: Using a large language model, the core text content of the specific information unit is vectorized to form a standardized query system that stores the structured standardized content and its corresponding vectorized representation.

[0011] The scope of the specification sets corresponding to each chapter of the preset design document: manually configure the mapping relationship between each chapter in the engineering design document and the corresponding one or at least two engineering specifications to form different specification sets corresponding to each chapter in the preset design document;

[0012] Vectorization and similarity query of user-selected text in a text editor: Select text in the text editor and use a large language model to vectorize the selected text to generate a user query vector and input it into the standard query system; the standard query system calculates the semantic similarity between the user query vector and the vectorized representation of each standard content, and then finds the vectorized representation that is most semantically similar to the user query vector in the standard set corresponding to the preset document section;

[0013] Providing the user with the corresponding standard content: through the user interface of the text editor add-on, combined with the preset document chapter corresponding standard set, the standard content corresponding to the vectorized representation content that is most semantically similar is displayed.

[0014] The engineering design document compliance checking method provided by this invention integrates optical character recognition (OCR) technology, a large language model (LLM), and regular expressions (RegEx) to preprocess massive PDF-formatted specification documents and construct a structured, vectorizable specification query system. When users compose design documents using word processing software, this specification query system, as a word processing software add-on, intelligently retrieves the most semantically similar specification entries, figures, tables, and other content from the specification query system based on the user's current editing content or selected text, and presents them in the user interface. This greatly facilitates staff in querying relevant specifications and indirectly assists in compliance assessments.

[0015] Preferably, in the step of text recognition and extraction of the specification document, the engineering specification document is a PDF document, and after being manually confirmed as the current valid version, optical character recognition technology is used to perform text recognition and extraction on the engineering specification document to be consulted.

[0016] Preferably, in the steps of checking and manually correcting text errors in the standardized text, the errors include text recognition errors, garbled characters, format anomalies or semantic incoherence generated during the optical character recognition process; the large language model highlights suspicious words or provides context-based correction suggestions.

[0017] Preferably, in the step of structural processing of the revised standard text, regular expressions are used to identify and extract the specific information units in the manually revised standard text according to predefined regular expression rules, including:

[0018] Item / clause content: Identify and extract each specific regulatory requirement, definition, or explanation;

[0019] Figure / Diagram Content: Identify and extract the titles and numbers of all diagrams in the standard text;

[0020] Table / Form Content: Identify and extract the titles, numbers, and key text content of all tables in the standard text;

[0021] PDF file location: records the file name and storage path of the original PDF document from which each extracted entry, figure, or table originates;

[0022] Page number information: records the specific page number of the entry, figure, or table in the original PDF document;

[0023] Item location: record the specific location of the item, figure, or table in the logical structure of the specification;

[0024] A structured specification data set is then obtained, wherein each entry / figure / table is accompanied by corresponding location data, and the location data includes physical location information and logical location information; the physical location information includes page number, paragraph number, line number and / or coordinate; the logical location information includes chapter, clause number and / or table number.

[0025] Preferably, in the steps of vectorizing the core text content and constructing the standardized query system, the core text content of each entry, figure, and table obtained after the structured processing is vectorized using a large language model, and the core text content includes the title and key description content;

[0026] Using a large language model to convert the unstructured text information of the core text content into a numerical vector in a high-dimensional space, so that semantically similar texts have a closer distance in the vector space;

[0027] The contents of the specification query system include:

[0028] Entry content: original, revised and structured standard text content;

[0029] Vectorized content: the numerical vector corresponding to the content of the entry;

[0030] Entry position: the logical position of the content in the specification system;

[0031] Original location: The physical location of the content in the original PDF document.

[0032] Preferably, in the step of vectorizing and similarity querying text selected by the user in a text editor, the text editor is WPS or WORD, the text is a text fragment being written or reviewed, and the text selected by the user is obtained through a WPS add-in or a WORD add-in;

[0033] The specification query system calculates the semantic similarity between the user query vector and the vectorized representation content of all the specification contents by calculating the cosine similarity between the two, and finds the vectorized representation content of the specification content that is most semantically similar, as well as its corresponding specification entries, figures, and table contents.

[0034] Preferably, in the step of providing the corresponding specification content to the user, auxiliary information is provided to the user during the engineering design document editing process, and the auxiliary information includes the chapter in which the current editing content is located and the specific selected text; the displayed specification content includes the entry content in the original specification text, the logical position of the entry in the specification text, and the physical position of the entry in the original PDF file.

[0035] Preferably, the engineering specification document is set as a specification, standard and / or legal and regulatory document related to water transport engineering, and the engineering design document is set as a water transport engineering design document.

[0036] In a second aspect, the present invention provides an electronic device, comprising:

[0037] at least one memory storing computer instructions;

[0038] at least one processor, which runs the computer instructions to perform the engineering design document compliance checking method according to the first aspect of the present invention;

[0039] Communication interface, used for exchanging data with external devices;

[0040] A communication bus is connected to the processor, the communication interface and the memory respectively, and is used to provide a data transmission path.

[0041] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the engineering design document compliance checking method described in the first aspect of the present invention.

[0042] The engineering design document compliance checking method, electronic device, and storage medium provided by the present invention have the following beneficial effects:

[0043] 1. The engineering design document compliance inspection method provided by the present invention has the following technical features and advantages:

[0044] First, the combination of OCR and LLM enables high-precision preprocessing of specification text. Unlike traditional text extraction that relies solely on OCR, this approach leverages OCR for efficient text extraction and incorporates a large language model to provide error warnings and assist in manual correction of initial OCR recognition results, ensuring the high accuracy, precision, and quality of specification text data. This effectively addresses the recognition errors often associated with traditional OCR when processing complex engineering specification documents (e.g., those containing numerous diagrams, tables, formulas, and special symbols), laying a solid data foundation for subsequent intelligent querying. This significantly improves the accuracy of the specification database and avoids query deviations caused by errors in the original data, a prerequisite for precise semantic querying.

[0045] Secondly, structured and vectorized storage: it uses regular expressions to structure standard text and extract key information; and uses a large language model to vectorize the content of each entry, figure, and table, building a semantically understandable vector database. This goes beyond the limitations of traditional keyword search and achieves a leap from keyword search to semantic search, that is, from "literal matching" to "semantic matching", greatly improving the intelligence and accuracy of queries.

[0046] Again, semantic similarity query: when the user selects text for query in the word processing tool, it vectorizes the text selected by the user and performs similarity matching in the vector space. By calculating the semantic similarity between vectors, it can find the most relevant normative content in terms of semantics. Even if the user uses imprecise keywords or different expressions, as long as the semantics are relevant, the normative query system can accurately match the corresponding normative entries, greatly improving the intelligence, accuracy and user experience of the query.

[0047] In addition, chapter-specification range preset: it effectively narrows the query space and improves query efficiency and relevance of results by manually presetting the specification range corresponding to the design document chapter.

[0048] In addition, a deep integration application model of word processing software add-ins and domain knowledge is proposed: the aforementioned OCR-LLM preprocessing and LLM semantic query capabilities are deeply integrated into the daily office environment used by water transport engineering designers in the form of WPS add-ins or Word add-ins. At the same time, the introduction of manually preset domain knowledge related to the design document chapters and specification scopes enables contextual awareness and accurate recommendations for queries. This tight integration and combination of domain knowledge enables the specification query system to be seamlessly embedded in the design workflow, providing immediate and relevant specification assistance, greatly improving the work efficiency and convenience of designers, and transforming specification query from a tedious independent task to an intelligent auxiliary function.

[0049] Finally, human-machine collaboration facilitates querying rather than mandatory judgment: The specification query system is positioned as an intelligent assistance tool—that is, it provides "assisted query" rather than "automated compliance checking," leaving the final decision on compliance to professional designers. This aligns with the professional and complex requirements of the engineering field. It fully leverages AI's strengths in data processing and semantic matching, while respecting and preserving the extremely high requirements for professional knowledge and human judgment in water transport engineering design. This leverages technology to improve efficiency while avoiding the risk of misjudgment that automated judgment can bring in complex engineering contexts. This system complements the strengths of AI and human expert intelligence, providing reliable intelligent assistance for high-precision, high-risk engineering design.

[0050] 2. Compared with the existing technology, the engineering design document compliance inspection method provided by the present invention can significantly improve the efficiency and accuracy of specification query, specifically including the following aspects:

[0051] First, it overcomes the drawback of inefficient manual review: Traditional manual review of specifications is extremely inefficient, time-consuming, and labor-intensive. This invention automates the previously labor-intensive search and comparison tasks through automated OCR, large language model-assisted correction, structured processing, and semantic vectorized query. Designers no longer need to page through PDFs or manually input keywords repeatedly. Simply select the document content, and the specification query system quickly presents relevant specifications, significantly saving time and labor costs and improving design and review efficiency.

[0052] Secondly, it improves the accuracy of standard queries: Traditional keyword searches are often limited to literal matching and are difficult to handle synonyms, near-synonyms, or different semantically related expressions. The present invention introduces a large language model for vectorization, realizing intelligent query based on semantic similarity. This means that even if the text selected by the user is different from the original standard expression, as long as the semantics are related, the standard query system can accurately match the most relevant standard entry, avoiding missed searches and wrong searches due to differences in expression, and significantly improving the accuracy of the query.

[0053] Thirdly, ensuring the quality of standard text: Through the collaboration of OCR and a large language model, the standard query system can indicate possible OCR recognition errors and, combined with manual corrections, ensure the high accuracy and quality of the data in the standard database. This provides a reliable foundation for subsequent intelligent queries and avoids query deviations caused by data source errors.

[0054] Finally, it has excellent expansion potential: As an auxiliary query tool, the structured specification library and vectorized index constructed by this specification query system have good scalability. Although currently positioned as a manual judgment system, it is easy to explore integration with more engineering design software or, based on the accuracy of basic data, gradually introduce some rule-based automatic inspection modules, thus facilitating the development of more advanced automated compliance review.

[0055] 3. The engineering design document compliance checking method provided by the present invention can solve the technical shortcomings of the existing technology in manually reviewing specification references in documents, such as low efficiency, low accuracy, inability to dynamically and effectively manage, and lack of uniformity; it can quickly, efficiently and accurately identify specification references (including names and numbers) in design documents and compare them with preset current specification information; it can realize semantic-level intelligent query through structured processing and vectorization of specification text, so that designers can quickly obtain specification entries related to the current work content; it can integrate the above functions into word processing software editors (such as WPS or WORD) to provide a convenient operation interface and feedback mechanism, significantly improving the efficiency of designers' specification query work when writing and reviewing documents; although the final compliance judgment still needs to be made manually, through automated identification, comparison and prompts, the difficulty and error rate of manual verification are greatly reduced, thereby indirectly improving the compliance of the referenced specifications in the document; it can improve the efficiency and accuracy of specification references in documents, facilitate the dynamic update management of specification content, and ensure the correctness of referenced specifications.

[0056] 4. The engineering design document compliance inspection method, electronic device, and storage medium provided by the present invention can achieve the following significant effects upon implementation:

[0057] Significantly improved efficiency in specification lookup: Through automated recognition, semantic search, and word processing software integration, designers can now find and locate specification items in seconds, down from minutes or even hours (in complex cross-reference scenarios). This significantly reduces the time designers spend searching for specifications during the design document writing process.

[0058] Improve designer job satisfaction: The tedious task of querying specifications is one of the pain points in designers' work. The application of this specification query system will greatly reduce this burden, allowing designers to focus more on innovation and solving technical problems, thereby improving their job satisfaction and sense of professional achievement.

[0059] Improve the level of enterprise knowledge management: The structured specification database and intelligent specification query system it constructs are actually important components of the enterprise specification knowledge management system, which helps to systematize and search the specification knowledge scattered in PDF documents, and enhance the value of the enterprise's knowledge assets. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a flow chart of a method for checking compliance of engineering design documents provided in one embodiment of the present invention.

[0061] Figure 2 It is a structural diagram of an electronic device provided by an embodiment of the present invention.

[0062] In the figure, 100 is a processor, 200 is a communication interface, 300 is a memory, and 400 is a communication bus. DETAILED DESCRIPTION

[0063] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] Example 1

[0065] like Figure 1 As shown, this embodiment provides a method for checking compliance of engineering design documents, including the following steps:

[0066] S11. Text recognition and extraction of specification documents: Use optical character recognition technology to recognize and extract text from the engineering specification documents to be consulted, forming editable and searchable specification texts;

[0067] S12. Error checking and manual correction of standard texts: A large language model is introduced to check and prompt errors in standard texts, and errors are manually corrected;

[0068] S13. Structural processing of the revised standard text: identifying and extracting specific information units in the standard text to obtain a structured standard data set, wherein each specific information unit in the standard data set is accompanied by corresponding physical location information and logical location information;

[0069] S14. Vectorization of core text content and establishment of a standardized query system: A large language model is used to vectorize the core text content of specific information units to form a standardized query system. The standardized query system stores the structured standardized content and its corresponding vectorized representation.

[0070] S21. Scope of specification sets corresponding to each chapter of the preset design document: manually configuring a mapping relationship between each chapter in the engineering design document and one or more corresponding engineering specifications to form different specification sets corresponding to each chapter in the preset design document;

[0071] S22. Vectorization and Similarity Query of User-Selected Text in the Text Editor: Select text in the text editor and use the large language model to vectorize the selected text to generate a user query vector, which is then input into the canonical query system. The canonical query system calculates the semantic similarity between the user query vector and the vectorized representations of each canonical content, and searches the canonical set corresponding to the pre-set document section for the vectorized representation that is most semantically similar to the user query vector.

[0072] S23. Providing the user with the corresponding standard content: Through the user interface of the text editor add-on, combined with the preset document chapter corresponding standard set, the standard content corresponding to the vectorized representation content that is most semantically similar is displayed.

[0073] The engineering design document compliance inspection method of this embodiment can be specifically divided into stages S1 and S2. Stage S1 is the specification library construction and preprocessing stage, and stage S2 is the word processing software add-in application and intelligent query stage. Stage S1 includes steps S11 to S14, whose goal is to convert massive amounts of PDF-formatted engineering specification documents into structured data that can be intelligently queried and understood by the specification query system. Stage S2 includes steps S21 to S23, which responds to user operations in the WPS or WORD editor in real time through the WPS or WORD software add-in and provides intelligent specification query assistance functions.

[0074] Specifically, during the text recognition and extraction process of the specification document in step S11, the engineering specification document is a PDF document, and after being manually confirmed as the current valid version, optical character recognition technology is used to perform text recognition and extraction on the engineering specification document to be consulted.

[0075] More specifically, in step S11, first, all engineering specification documents stored in PDF format that need to be consulted are manually confirmed to be the current valid versions to ensure the basic accuracy of the knowledge base; then these engineering specification documents are imported into the specification query system;

[0076] Secondly, optical character recognition technology is used to identify and extract text from these PDF documents. Optical character recognition technology can convert the text content in the form of images in PDF documents into an editable and searchable text format. At this time, the goal of the standard query system is to obtain the text information of the original PDF document as completely as possible.

[0077] Then, preliminary recognition is performed to obtain the standard text.

[0078] During the text error checking and manual correction process of the standard text in step S12, errors include text recognition errors, garbled characters, format anomalies, or semantic incoherence generated during the optical character recognition process.

[0079] More specifically, in step S12, the initially recognized standard text generated by the large language model in step S12 is firstly introduced for error checking. The large language model, through its language understanding capabilities, identifies text recognition errors, garbled characters, formatting anomalies, or semantic incoherence that may occur during the OCR process, and indicates these potential errors, such as highlighting suspicious words or providing context-based correction suggestions.

[0080] Then, manual corrections are made to the errors suggested by the large language model, thereby ensuring the high accuracy of the standardized text. Through human-computer collaboration, the shortcomings of single OCR technology in recognizing complex documents (such as standardized text containing a large number of charts, formulas, and special symbols) are made up, greatly improving the quality of text data.

[0081] During the structural processing of the revised standard text in step S13, regular expressions are used to identify and extract specific information units in the manually revised standard text according to predefined regular expression rules, including:

[0082] Item / clause content: Identify and extract each specific regulatory requirement, definition, or explanation;

[0083] Figure / Diagram Content: Identify and extract the titles and numbers of all diagrams in the standard text;

[0084] Table / Form Content: Identify and extract the titles, numbers, and key text content of all tables in the standard text;

[0085] PDF file location: records the file name and storage path of the original PDF document from which each extracted entry, figure, or table originates;

[0086] Page number information: records the specific page number of the entry, figure, or table in the original PDF document;

[0087] Item location: record the specific location of the item, figure, or table in the logical structure of the specification;

[0088] Then, a structured normative data set is obtained, in which each entry / figure / table is accompanied by corresponding location data, which includes physical location information and logical location information; the physical location information includes page number, paragraph number, line number and / or coordinate; the logical location information includes chapter, clause number and / or table number.

[0089] The coordinates in the physical location information may be pixel coordinates in a PDF or an image.

[0090] During the vectorization processing of the core text content and the construction of the standardized query system in step S14, the core text content of each entry, figure, and table obtained after the structural processing is vectorized using a large language model. The core text content includes the title and key description content.

[0091] Using a large language model, the unstructured text information of the core text content is converted into numerical vectors in a high-dimensional space, so that semantically similar texts have closer distances in the vector space;

[0092] The contents of the standard query system include:

[0093] Entry content: original, revised and structured standard text content;

[0094] Vectorized content: the numerical vector corresponding to the content of the entry;

[0095] Entry position: the logical position of the content in the specification system;

[0096] Original location: The physical location of the content in the original PDF document.

[0097] More specifically, in step S14, the core text content is vectorized using a large language model (LLM). Vectorization converts unstructured text information into numerical vectors in a high-dimensional space (i.e., embeddings), so that semantically similar texts are closer in the vector space. In this way, the semantic similarity of texts can be determined by calculating the distance between vectors.

[0098] A canonical query system can also be called a vector database, where:

[0099] Vectorized content: The numeric vector corresponding to the content of this entry, i.e., the embedding.

[0100] Entry location: The logical location of the content in the specification system; for example, in which chapter of which specification.

[0101] Original location: The physical location of the content in the original PDF file, that is, the standard file path, corresponding to which page in the PDF.

[0102] Specifically, in step S21, the scope of regulations corresponding to the preset design document sections is manually confirmed. That is, before the specification query system is deployed and used, the scope of regulations potentially covered by each major section (chapter) in the engineering design document is pre-established through manual confirmation. For example, in the case of water transport engineering design documents, the "Preliminary Design Document, Chapter 2, Natural Conditions, Section 3, Hydrology" section is associated with a series of relevant regulations, including the "Seaport Master Design Specification," the "River Port Master Design Specification," the "Port and Waterway Hydrology Specification," and the "Inland Waterway Navigation Standard." This pre-setting mechanism, as a form of domain knowledge, significantly narrows the scope of subsequent queries, improving query efficiency and result accuracy, avoiding ineffective searches within irrelevant specifications, and thus optimizing the user experience.

[0103] More specifically, in step S21, after the specification is updated, the manually configured preset specification set is also updated.

[0104] In step S22, during the vectorization and similarity query of text selected by the user in the text editor, the text editor is WPS or Word, and the text is a text fragment being written or reviewed. For example, the text fragment being written or reviewed can be a phrase, a sentence, or a description. The text selected by the user is then obtained through the WPS add-in or the Word add-in.

[0105] The specification query system calculates the semantic similarity between the user query vector and the vectorized representation of all specification contents by calculating the cosine similarity between the two, and finds the vectorized representation of the semantically most similar specification content, as well as its corresponding specification entries, figures, and tables.

[0106] This semantic-based query overcomes the limitations of traditional keyword searches. Even if the expressions are different, as long as the semantics are related, they can be accurately matched.

[0107] In the process of providing the corresponding specification content to the user in step S23, auxiliary information is provided to the user during the engineering design document editing process, and the auxiliary information includes the chapter in which the current editing content is located and the specific selected text; the displayed specification content includes the entry content in the original specification text, the logical position of the entry in the specification text, and the physical position of the entry in the original PDF file.

[0108] More specifically, in step S23, the chapter in which the currently edited content is located is provided in combination with the preset specification scope of step S21; based on the semantic similarity query result of step S22, the specific selected text is provided; and the most semantically relevant specification content found is displayed to the staff through the user interface of the WPS add-in or WORD add-in. These contents may include: the original specification entry content, the logical position of the entry in the specification (for example, Chapter X, Article Y), and the specific page number of the entry in the original PDF file, etc.; this makes it easier for staff to query and refer to relevant specification content, which serves as an important reference for their design, writing, and manual compliance judgment; although it does not perform automated compliance judgment, it provides efficient and accurate specification information support to assist humans in making correct compliance decisions.

[0109] Specifically, the engineering specification documents are set to documents such as specifications, standards and / or laws and regulations related to water transport engineering, and the engineering design documents are set to water transport engineering design documents.

[0110] Example 2

[0111] like Figure 2 As shown, this embodiment provides an electronic device, including:

[0112] at least one memory 300 storing computer instructions;

[0113] At least one processor 100, which runs computer instructions to perform the engineering design document compliance checking method according to the first embodiment;

[0114] Communication interface 200, used for exchanging data with external devices;

[0115] The communication bus 400 is connected to the processor 100 , the communication interface 200 , and the memory 300 , respectively, and is used to provide a data transmission path.

[0116] Processor 100 (Processor) can be hardware configured to perform computing tasks, such as a central processing unit (CPU) or an application-specific integrated circuit (ASIC). It is responsible for executing the various steps of the method of Example 1, including core computing tasks such as OCR processing, large language model inference, regular expression matching, vectorized computing, and database queries.

[0117] Communication interface 200 is used to exchange data with other devices (such as the computer hosting the WPS editor, the regulatory document storage server, and internet services). For example, data transmission is achieved through the network interface module, ensuring that the regulatory query system can access regulatory documents, update data, and communicate with word processing tools.

[0118] Memory 300 (Memory) is used to store program instructions, data to be processed, processed standard text, structured standard database, vector database, user query history, and data required for the operation of WPS add-ins or Word add-ins, etc. It can be random access memory (RAM), read-only memory (ROM), solid-state drive (SSD) or mechanical hard disk, etc.

[0119] The communication bus 400 is used to connect various hardware components such as the processor 100, the communication interface 200, and the memory 300, and to provide a data transmission path to ensure that the modules within the engineering design document compliance inspection device can efficiently exchange data and work together.

[0120] The processor 100 , the communication interface 200 , and the memory 300 communicate with each other via the communication bus 400 .

[0121] Example 3

[0122] A computer-readable storage medium stores a computer program, which, when executed by the processor 100, implements the engineering design document compliance checking method of embodiment 1.

[0123] The engineering design document compliance inspection method, electronic device, and storage medium of the above-mentioned embodiments are primarily applied to the field of water transport engineering design. However, their core technical concepts and methods are universal and can be extended to other professional fields that require frequent reference to and citation of complex regulatory systems, such as:

[0124] Engineering design industry: including various engineering design fields such as water conservancy and hydropower, construction, transportation (highway, railway), municipal administration, and electricity. These fields have a huge and constantly updated system of specifications and standards.

[0125] Scientific research and academic fields: Assist scientific researchers, university teachers and students to quickly access and cite relevant standards and specifications when writing papers and research reports.

[0126] Legal and regulatory inquiry: In the legal, financial and other industries, it assists professionals in efficiently finding and understanding complex legal provisions, rules and regulations.

[0127] Quality management and auditing: used in the company's internal quality management system to ensure that products or services comply with relevant standards and regulations.

[0128] The engineering design document compliance checking method, electronic device, and storage medium of the above-mentioned embodiments can be widely applied to the following specific aspects in the field of water transport engineering design:

[0129] Preliminary design stage: When preparing preliminary design documents, designers quickly query relevant specifications for various basic conditions (such as hydrology, geology, and meteorology) as well as overall layout, major structures, etc.

[0130] Construction drawing design stage: When conducting detailed structural calculations, material selection, equipment configuration, etc., accurately locate specific design, construction and acceptance specifications.

[0131] Design review and proofreading: When designers conduct self-review or mutual proofreading, they can use the specification query system to assist in checking whether the specifications referenced in the document are correct and meet the latest requirements.

[0132] Technical communication and training: A tool for new designers or trainers to quickly familiarize themselves with industry standards.

[0133] Among them, "specification update" refers to the version revision or content change of engineering specifications (such as national standards and industry standards).

[0134] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0136] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0137] These computer program instructions can also be loaded onto a computer or other programmable computer device so that a series of operating steps are executed on the computer or other programmable computer device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable computer device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0138] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for checking compliance of engineering design documents, characterized in that: The following steps are involved: Text recognition and extraction of specification documents: Use optical character recognition technology to recognize and extract text from engineering specification documents to be consulted, forming editable and searchable specification texts; Error checking and manual correction of standard texts: A large language model is introduced to check and prompt errors in the standard texts, and the errors are manually corrected; Structural processing of the revised standard text: identifying and extracting specific information units in the standard text to obtain a structured standard data set, wherein each specific information unit in the standard data set is accompanied by corresponding physical location information and logical location information; Vectorization of core text content and establishment of a standardized query system: Using a large language model, the core text content of the specific information unit is vectorized to form a standardized query system that stores the structured standardized content and its corresponding vectorized representation. The scope of the specification sets corresponding to each chapter of the preset design document: manually configure the mapping relationship between each chapter in the engineering design document and the corresponding one or at least two engineering specifications to form different specification sets corresponding to each chapter in the preset design document; Vectorization and similarity query of user-selected text in a text editor: Select text in the text editor and use a large language model to vectorize the selected text to generate a user query vector and input it into the standard query system; the standard query system calculates the semantic similarity between the user query vector and the vectorized representation of each standard content, and then finds the vectorized representation that is most semantically similar to the user query vector in the standard set corresponding to the preset document section; Providing the user with the corresponding standard content: through the user interface of the text editor add-on, combined with the preset document chapter corresponding standard set, the standard content corresponding to the vectorized representation content that is most semantically similar is displayed.

2. The method for checking compliance of engineering design documents according to claim 1, characterized in that: In the step of text recognition and extraction of specification documents, the engineering specification document is a PDF document, and after being manually confirmed as the current valid version, optical character recognition technology is used to perform text recognition and extraction on the engineering specification document to be consulted.

3. The method for checking compliance of engineering design documents according to claim 2, characterized in that: In the step of checking and manually correcting text errors in the standard text, the errors include text recognition errors, garbled characters, abnormal formats, or semantic incoherence generated during the optical character recognition process; Large language models highlight suspicious words or provide correction suggestions based on the context.

4. The method for checking compliance of engineering design documents according to claim 3, characterized in that: In the step of structural processing of the revised standard text, regular expressions are used to identify and extract the specific information units in the manually revised standard text according to predefined regular expression rules, including: Item / clause content: Identify and extract each specific regulatory requirement, definition, or explanation; Figure / Diagram Content: Identify and extract the titles and numbers of all diagrams in the standard text; Table / Form Content: Identify and extract the titles, numbers, and key text content of all tables in the standard text; PDF file location: records the file name and storage path of the original PDF document from which each extracted entry, figure, or table originates; Page number information: records the specific page number of the entry, figure, or table in the original PDF document; Item location: record the specific location of the item, figure, or table in the logical structure of the specification; A structured specification data set is then obtained, wherein each entry / figure / table is accompanied by corresponding location data, and the location data includes physical location information and logical location information; the physical location information includes page number, paragraph number, line number and / or coordinate; the logical location information includes chapter, clause number and / or table number.

5. The method for checking compliance of engineering design documents according to claim 4, characterized in that: In the steps of vectorizing the core text content and building a standardized query system, the core text content of each entry, figure, and table obtained after structured processing is vectorized using a large language model, the core text content including the title and key description content; Using a large language model to convert the unstructured text information of the core text content into a numerical vector in a high-dimensional space, so that semantically similar texts have a closer distance in the vector space; The contents of the specification query system include: Entry content: original, revised and structured standard text content; Vectorized content: the numerical vector corresponding to the content of the entry; Entry position: the logical position of the content in the specification system; Original location: The physical location of the content in the original PDF document.

6. The method for checking compliance of engineering design documents according to claim 5, characterized in that: In the step of vectorizing and similarity querying text selected by the user in a text editor, the text editor is WPS or WORD, the text is a text fragment being written or reviewed, and the text selected by the user is obtained through a WPS add-in or a WORD add-in; The specification query system calculates the semantic similarity between the user query vector and the vectorized representation content of all the specification contents by calculating the cosine similarity between the two, and finds the vectorized representation content of the specification content that is most semantically similar, as well as its corresponding specification entries, figures, and table contents.

7. The method for checking compliance of engineering design documents according to claim 6, characterized in that: In the step of providing the corresponding specification content to the user, auxiliary information is provided to the user during the engineering design document editing process, and the auxiliary information includes the chapter in which the current editing content is located and the specific selected text; the displayed specification content includes the entry content in the original specification text, the logical position of the entry in the specification text, and the physical position of the entry in the original PDF file.

8. The method for checking compliance of engineering design documents according to any one of claims 1 to 7, characterized in that: The engineering specification document is set as a specification, standard and / or legal and regulatory document related to water transport engineering, and the engineering design document is set as a water transport engineering design document.

9. An electronic device, characterized in that: include: at least one memory storing computer instructions; at least one processor, which runs the computer instructions to perform the engineering design document compliance checking method according to any one of claims 1 to 8; Communication interface, used for exchanging data with external devices; A communication bus is connected to the processor, the communication interface and the memory respectively, and is used to provide a data transmission path.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method for checking compliance of engineering design documents according to any one of claims 1 to 8 is implemented.

Images