A document translation method and device, electronic equipment and storage medium

Abstract

This invention discloses a document translation method, apparatus, electronic device, and storage medium, relating to the field of translation technology. The method includes: parsing the document to be translated to determine the content to be translated; breaking down the content into translatable units; determining the backfilling location information for each translatable unit; identifying protected objects within each translatable unit; replacing the protected objects with target placeholders to obtain target translation units; establishing a mapping relationship between protected objects and target placeholders; generating a set of terminology constraints based on the text content of the document to be translated and the target translation requirements; integrating the target translation units and the set of terminology constraints to construct a translation task; generating an original translation using a preset translation model; restoring the protected objects to the original translation according to the mapping relationship to obtain a translation to be verified; performing a consistency check on the translation to be verified and generating a verification result; generating the target translation based on the verification result and backfilling location information, thereby improving the accuracy of document translation.

Description

Technical Field

[0001] This invention relates to the field of translation technology, and in particular to a document translation method, apparatus, electronic device, and storage medium. Background Technology

[0002] In the context of global wind power engineering R&D and collaboration, technical documents such as wind power technical drawings, technical specifications, test reports, and engineering descriptions generally require cross-language translation between Chinese and English. This places extremely high demands on the professionalism, terminology consistency, content fidelity, and format completeness of the translations. The wind power field has a large number of proprietary equipment models, process parameters, component names, test standards, and industry-specific expressions. The terminology is highly specialized, the expression standards are stringent, and updates are frequent, making it difficult for general translation solutions to adapt to these unique translation scenarios.

[0003] In routine wind power translation, traditional machine translation methods cannot guarantee consistent translation of the same technical terms throughout the entire document, severely impacting the professionalism and readability of the document. General translation models also perform poorly in wind power contexts, lacking domain-specific knowledge and resulting in translations that fail to meet professional requirements. Furthermore, traditional systems struggle to quickly adapt to the emergence of new technical terms and expressions, leading to translations that may not reflect the latest technological advancements. Therefore, this invention relates to the field of intelligent translation technology, specifically a fully automated intelligent translation system for wind power technical document translation. Its main purpose is to address issues such as inconsistent terminology translation and low efficiency in wind power translation, achieving a closed-loop operation of wind power terminology mining, usage, and iteration. In summary, existing wind power document translation technologies suffer from technical pain points such as poor terminology consistency, low professional adaptability, high costs of manual review, and uncontrollable translation quality, necessitating an automated intelligent translation method adapted to wind power technical document scenarios. Summary of the Invention

[0004] This invention provides a document translation method, apparatus, electronic device, and storage medium to solve the problems of poor consistency in terminology translation, low professional adaptability, high cost of manual review, and uncontrollable translation quality in existing wind power document translation technologies.

[0005] According to one aspect of the present invention, a document translation method is provided, wherein the method includes: The document to be translated is parsed to determine the content to be translated in the document, the content to be translated is split into at least two translatable units, and the backfilling positioning information of each translatable unit is determined. The protected object in each of the translatable units is determined, and the target translation unit is obtained by replacing the protected object with the target placeholder. A mapping relationship between the protected object and the target placeholder is established. A set of terminology constraints is generated based on the text content of the document to be translated and the target translation requirements. The target translation unit and the set of terminology constraints are then integrated to construct a translation task. The translation task is input into a preset translation model, and the original translation corresponding to the translation task is generated by the preset translation model. The protected object is restored to the original translation according to the mapping relationship to obtain the translation to be verified. Perform a consistency check on the translation to be checked and generate a check result. Generate the target translation based on the check result and the backfilling positioning information.

[0006] According to another aspect of the present invention, a document translation apparatus is provided, wherein the apparatus comprises: The content extraction module is used to perform document parsing on the document to be translated, determine the content to be translated in the document to be translated, split the content to be translated into at least two translatable units, and determine the backfilling positioning information of each translatable unit; The relationship determination module is used to determine the protected object in each of the translatable units, obtain the target translation unit by replacing the protected object with the target placeholder, and establish a mapping relationship between the protected object and the target placeholder; The task construction module is used to generate a set of terminology constraints based on the text content of the document to be translated and the target translation requirements, and to integrate the target translation unit and the set of terminology constraints to construct a translation task; An initial translation module is used to input the translation task into a preset translation model, generate the original translation corresponding to the translation task through the preset translation model, and restore the protected object to the original translation according to the mapping relationship to obtain the translation to be verified. The translation generation module is used to perform consistency verification on the translation to be verified and generate verification results, and generate target translation based on the verification results and the backfill positioning information.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the document translation method according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the document translation method according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, embodiments of the present invention also provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the document translation method of any embodiment of the present invention.

[0010] The technical solution of this invention involves parsing the document to be translated to determine the content to be translated, splitting the content into at least two translatable units, determining the backfilling location information of each translatable unit, identifying the protected object in each translatable unit, replacing the protected object with a target placeholder to obtain the target translation unit, establishing a mapping relationship between the protected object and the target placeholder, generating a terminology constraint set based on the text content of the document to be translated and the target translation requirements, integrating the target translation unit and the terminology constraint set to construct a translation task, inputting the translation task into a preset translation model, generating the original translation corresponding to the translation task through the preset translation model, restoring the protected object to the original translation according to the mapping relationship to obtain the translation to be verified, performing a consistency check on the translation to be verified and generating a check result, and generating the target translation based on the check result and backfilling location information. This reduces the risk of mistranslation of professional terms, multiple translations of the same term, and errors in translation backfilling, and improves the accuracy, proofreadability, and traceability of engineering and technical document translations.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a document translation method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a document translation method provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of a document translation device according to Embodiment 4 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the document translation method of this invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1 This is a flowchart of a document translation method according to Embodiment 1 of the present invention. This embodiment is applicable to the automated translation of documents requiring translation in the wind power field. The method can be executed by a document translation device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes: S110. Perform document parsing on the document to be translated, determine the content to be translated in the document, split the content to be translated into at least two translatable units, and determine the backfilling location information for each translatable unit.

[0017] The "documents to be translated" refers to technical documents in the field of wind power engineering. Generally, these documents are in Word or OOXML formats. They can include, but are not limited to, wind turbine technical specifications, engineering design reports, testing and inspection documents, construction process documents, and equipment parameter manuals. The content library of these documents includes main text paragraphs, tables, formulas, images, approval pages, tables of contents, and titles. The content to be translated is the valid text within the document that needs language conversion, primarily consisting of translatable semantic content such as the table of contents, main text, and table cell text. A translatable unit refers to the smallest independent translation processing unit obtained after breaking down the content to be translated. Generally, translatable units are broken down using document paragraphs and table cells as boundaries, and there are multiple translatable units. Backfilling positioning information refers to the information binding the order and position of each translatable unit.

[0018] In this embodiment, a document to be translated can be loaded, and document parsing can be performed to identify structural information such as the cover, approval page, revision page, table of contents, and main text. By using preset content recognition rules, translatable text content is distinguished from untranslatable fixed content, and the content requiring language conversion is selected. In one embodiment, the content to be translated may include a table of contents and main text. Then, based on document paragraph boundaries and independent semantic boundaries of table cells, the continuous content to be translated is split into multiple independent translatable units. During actual operation, during the splitting process, information such as the location coordinates, chapter affiliation, paragraph number, cell row and column identifiers, and title number format of each translatable unit is simultaneously collected to generate backfill positioning information corresponding to each translatable unit.

[0019] S120. Determine the protected object in each translatable unit, obtain the target translation unit by replacing the protected object with the target placeholder, and establish the mapping relationship between the protected object and the target placeholder.

[0020] The protected object refers to fixed content within the translatable unit that is specific to the wind power sector, prohibited from translation and tampering, and can be preset identifiers. For example, the protected object may include, but is not limited to, core fixed information such as equipment model, drawing number, industry standard number, calculation formula, precise numerical values, units of measurement, fixed table fields, and image-related identifiers. The target placeholder is an automatically generated, unique, and semantically unambiguous temporary character identifier used to temporarily replace the protected object. The mapping relationship stores information about the structured relationship between protected objects and target placeholders; the mapping relationship can store matching data between each target placeholder and its corresponding protected object.

[0021] In this embodiment, preset wind power-specific identification rules can be invoked, combined with regular expression matching and keyword precision retrieval technology, to scan and identify the text content of each translatable unit one by one, determining all protected objects contained in each translatable unit, such as equipment model, standard number, numerical parameters, and unit of measurement. A corresponding target placeholder is assigned to each identified independent protected object. The protected objects within the translatable unit are replaced one by one with the corresponding target placeholder, eliminating all untranslatable protected objects in the translatable content, completing text standardization processing to obtain the target translation unit. Simultaneously, the correspondence between each set of protected objects and target placeholders is recorded, constructing a structured mapping relationship. In one embodiment, the mapping relationship can be in the form of a table or a document.

[0022] S130. Generate a set of terminology constraints based on the text content of the document to be translated and the target translation requirements, integrate the target translation unit and the set of terminology constraints, and construct the translation task.

[0023] The target translation unit refers to the translation unit obtained by replacing all protected objects within the translatable unit with target placeholders. The target translation requirements refer to the customized requirements parameters for this document translation task. For example, target translation requirements may include task constraints such as the translation language, document type, terminology standardization requirements, comparative formatting requirements, and highlighting requirements. The terminology constraint set is a set of exclusive terminology rules adapted to the current wind power document scenario. It is generated based on the textual characteristics and translation requirements of the document to be translated and includes constraints such as wind power professional standard terminology, standard translations, prohibited translations, and applicable scenarios for terminology, used to constrain the output of the preset translation model. The translation task refers to a structured, encapsulated inference task data package. Generally, a translation task includes the target translation unit and the terminology constraint set.

[0024] In this embodiment, textual features and domain attributes of the document to be translated can be extracted. Combined with the target translation requirements of the document, a pre-defined terminology resource library is retrieved, and terminology constraints such as professional terminology, standard translations, and prohibited translations suitable for the document are matched to generate a terminology constraint set. Subsequently, the target translation unit corresponding to each translatable unit and the generated terminology constraint set are structurally encapsulated and associated to form a translation task.

[0025] S140. Input the translation task into the preset translation model, generate the original translation corresponding to the translation task through the preset translation model, and restore the protected object to the original translation according to the mapping relationship to obtain the translation to be verified.

[0026] The preset translation model can be a pre-trained translation model. In actual operation, the preset translation model can be an artificial intelligence model with text translation capabilities in the wind power field, such as a Transformer architecture machine translation model or a large language model. The original translation is the initial translation output by the preset translation model after reading the translation task and reasoning according to the terminology constraint rules, and all target placeholders are still retained in the text. The translation to be verified refers to the complete initial translation text obtained after replacing all target placeholders in the original translation with the protected objects according to the preset mapping relationship.

[0027] In this embodiment, the translation task can be input into a preset translation model. The preset translation model reads the target translation unit text and terminology constraint set within the translation task, and outputs the original translation that conforms to professional standards and corresponds to the translation task. Then, a pre-stored mapping relationship is retrieved, and all target placeholders in the original translation are traversed, matching the protected object corresponding to each target placeholder one by one. Using a one-to-one replacement method, all target placeholders are restored to the initial protected object content, resulting in the translation to be verified.

[0028] S150. Perform consistency verification on the translation to be verified and generate the verification result. Generate the target translation based on the verification result and the backfill location information.

[0029] Consistency verification can be a compliance check process for the translation to be verified. The dimensions of consistency verification can include consistency in terminology translation, integrity of the restored protected objects, and integrity of text content. It is used to identify problems such as incorrect terminology, missing content, or abnormal alteration in the translation. The target translation refers to the translated document after consistency verification, terminology highlighting, and in-situ comparison and backfilling. The target translation can retain the entire structure, format, and style of the original document, enabling a side-by-side display of the two languages.

[0030] In this embodiment, consistency verification can be performed on the translation to be verified through a set of terminology constraints and mapping relationships. For example, it can check whether the translation of wind power terminology is compliant and consistent, whether the protected objects are fully restored, and whether there are any anomalies such as omissions, mistranslations, or missing information in the text. All anomalies identified during the verification process are summarized, generating a verification result containing anomaly entries, locations, and types, and corresponding correction suggestions are generated simultaneously. At the same time, wind power terminology in the translation to be verified is automatically identified and highlighted for easy manual review later. Finally, based on the backfilling location information corresponding to each target translation unit, the processed translation to be verified is backfilled to the corresponding position in the document to be translated, with the original text and translation arranged in a comparative format. The original document object is reused throughout the process, preserving the original document structure and formatting, ultimately generating a target translation with complete format and standardized terminology. The target translation, verification result, and correction suggestions are then output in a correlated manner.

[0031] In this embodiment of the invention, document parsing is performed on the document to be translated to determine the content to be translated in the document. The content to be translated is split into at least two translatable units, and the backfilling location information of each translatable unit is determined. The protected object in each translatable unit is determined, and the target translation unit is obtained by replacing the protected object with a target placeholder. A mapping relationship between the protected object and the target placeholder is established. A set of terminology constraints is generated according to the text content of the document to be translated and the target translation requirements. The target translation unit and the set of terminology constraints are integrated to construct a translation task. The translation task is input into a preset translation model, and the original translation corresponding to the translation task is generated through the preset translation model. The protected object is restored to the original translation according to the mapping relationship to obtain the translation to be verified. A consistency check is performed on the translation to be verified and a check result is generated. The target translation is generated according to the check result and the backfilling location information. This reduces the risk of mistranslation of professional terms, multiple translations of the same term, and errors in translation backfilling, and improves the accuracy, proofreadability, and traceability of engineering and technical document translations.

[0032] In one embodiment, after performing a consistency check on the translation to be checked and generating a check result, and generating the target translation based on the check result and backfilling location information, the method further includes: Collect the terminology constraints corresponding to the abnormal information, determine the modification records corresponding to the abnormal information, and generate a review feedback log by combining the terminology constraints and modification records. Extract candidate terms from the review feedback log and generate candidate term update suggestions that include the original text of the candidate term, the candidate translation, the frequency of occurrence, the context, the manual modification record, and the recommendation reason; The updated candidate terms suggestions are pushed to the review panel, which then determines whether the candidate terms are the target terms. When a candidate term is determined to be the target term, the candidate term is stored in a preset term resource library, and the generation rules of the term constraint are updated according to the applicable scenarios of the candidate term.

[0033] Among them, "abnormal information" refers to various abnormal information identified by the consistency check of the translation to be verified. For example, abnormal information may include verification alarms such as terminology translation not conforming to the terminology constraint set, inconsistencies in terminology translations, and non-standard professional expressions. Abnormal information can record the abnormal location, conflicting terms, the current translation, and the constraint standard translation. Terminology constraints refer to the rules governing the translation task execution process. For example, belonging constraints can include the original professional terminology, standard translation, prohibited translations, and applicable scenarios. Terminology constraints can be derived from a pre-set terminology resource library. Modification records refer to the operation records generated by manual reviewers when adjusting the translation to be verified, including the original translation, the modified translation, the modification location, the modification time, and the reviewer's comments. Review feedback logs refer to structured persistent logs that can integrate verification abnormal data, effective terminology constraints, and manual modification records. Candidate terms refer to professional expressions automatically mined from the review feedback logs that have not yet been included in the pre-set terminology resource library. Generally, candidate terms may have translation disputes, be frequently manually corrected, or continuously trigger terminology inconsistency anomalies, requiring manual determination for inclusion in the library. The candidate term update suggestion refers to a suggested data package that summarizes all statistical information of candidate terms. The review interface can be a human-computer interaction interface for terminology administrators or reviewers, supporting operations such as viewing candidate terminology information, performing confirmation, modification, and merging. Target terms refer to candidate terms that have been reviewed and confirmed and are allowed to be formally included in the preset terminology resource library. The preset terminology resource library is a database that persistently stores professional standard terms. Generally, it can store information such as the original terminology text, standard translation, applicable scenarios, prohibited translations, version information, and change records, used for subsequent generation of terminology constraint sets. The terminology constraint generation rules are the matching logic for filtering information from the preset terminology resource library and constructing a terminology constraint set specific to a single translation task. In practical applications, the terminology constraint generation rules can include information such as the applicable document type, chapter scenario, and contextual constraints.

[0034] In this embodiment, anomaly information from consistency verification output can be collected and associated with the terminology constraints used in the translation task. Simultaneously, manual review operations corresponding to the same translatable unit are retrieved, and complete modification records are extracted. Information such as anomaly details, associated terminology constraints, original text, original translation, text before and after modification, review comments, document type, and chapter position are linked and integrated, and structuredly written into the review feedback log. The review feedback log is traversed, and candidate terms are selected according to a preset mining strategy, such as expressions that frequently trigger terminology inconsistency anomalies, professional terms that have been repeatedly manually corrected to fixed translations, and wind power professional expressions that appear repeatedly but are not covered by the terminology database. The mined candidate terms are merged, deduplicated, and their frequency is counted. Multiple real context samples are extracted and bound to corresponding historical manual modification records. Recommendation reasons are generated based on anomaly frequency and manual correction status, and all information is integrated to form candidate terminology update suggestions. If multiple different translations exist for the same candidate term, multiple candidate translations and their corresponding contexts are retained simultaneously, without automatic selection of the best translation. The system batch pushes updated candidate terms to the review panel for display. Terminology administrators can view the original text, candidate translation, frequency of occurrence, engineering context, historical review and modification records, and reasons for recommendation for each candidate term, comprehensively judging whether the professional expression is suitable for inclusion in the pre-defined terminology resource library. If the review panel determines that the candidate term is the target term, the original text, standard translation, applicable scenarios, prohibited translations, and review information are written to the pre-defined terminology resource library. The system can parse the applicable document type, engineering scenario, and context scope corresponding to the candidate term, updating the terminology constraint generation rules. During subsequent document translation and terminology constraint set generation, the system automatically matches and loads the newly added term according to the updated rules based on the attributes of the document to be translated, constraining the model output during the translation phase to ensure consistency of translated terminology and improve the timeliness of the pre-defined terminology resource library.

[0035] Example 2 Figure 2 This is a flowchart of a document translation method according to Embodiment 2 of the present invention. This embodiment is a further optimization and extension based on the above embodiments, and can be combined with various optional technical solutions in the above embodiments. Figure 2 As shown, the method includes: S201. Identify the different layout structures within the document to be translated, and determine the content to be translated according to the layout structure.

[0036] Among them, layout structure refers to the different content layout structure types in the document to be translated, mainly including ordinary paragraphs in the main text, multi-level headings, various tables, table of contents, headers and footers, and fixed identifier pages.

[0037] In this embodiment, document structure analysis technology can be used to identify various internal layout structures of the document, including the cover, image paragraphs, table of contents lines, and revision pages. Translable and non-translatable content are then categorized based on these different layout structures. Generally, the layout structures requiring translation can be pre-defined. In one embodiment, non-translatable content can be protected.

[0038] S202. The content to be translated is split into complete paragraphs and complete cells to obtain at least two translatable units.

[0039] In this embodiment, the content to be translated can be split into multiple translatable units by paragraphs or complete cells.

[0040] S203. Collect the position and sequence information of each translatable unit in the document to be translated, and generate corresponding backfill positioning information according to the position and sequence information.

[0041] The location information is structured data used to locate the spatial position of translatable units within the document to be translated, including information such as chapter affiliation, paragraph number, table row and column coordinates, hierarchical index, and page position. The sequence information records the natural layout order of each translatable unit within the document to be translated.

[0042] In this embodiment, location data such as the document chapter, paragraph number, table row and column coordinates, hierarchical index, and page position of each translatable unit can be collected as location information. The order information of the documents to be translated is determined, and the collected location information and the order information are structurally integrated and associated to generate backfill location information corresponding to each translatable unit.

[0043] S204. Identify the preset identifier content contained in each translatable unit and treat the preset identifier content as the object of protection.

[0044] Among them, the preset identification content refers to the exclusive fixed content in the project document that is prohibited from translation and tampering, which is domain-specific fixed information.

[0045] In this embodiment, the text content of each translatable unit can be traversed using a combination of regular expression matching and keyword retrieval to identify various preset identifiers contained in the text, such as wind power equipment models, engineering drawing numbers, industry standard numbers, technical parameter values, units of measurement, and fixed fields. The identified preset identifiers are then marked as protected objects.

[0046] S205. Assign target placeholders to each protected object, replace the corresponding protected object in the translatable unit with the target placeholders, and generate the processed target translation unit.

[0047] In this embodiment, the corresponding target placeholder can be withdrawn for each protected object, and the corresponding protected object in the translatable unit can be replaced by the target placeholder to obtain the target translation unit.

[0048] S206. Record the correspondence between the protected object and the corresponding target placeholder to form a mapping relationship.

[0049] In this embodiment, a mapping table can be constructed that corresponds to the protected object and the corresponding target placeholder.

[0050] S207. Extract text feature information based on the text content of the document to be translated, and determine the target translation requirements of the document to be translated.

[0051] Among them, text feature information refers to the text attribute features of the document to be translated, including information such as the wind power field scenario to which the document belongs, text semantic type, distribution of professional vocabulary, sentence structure, and document format type.

[0052] In this embodiment, the text content of the document to be translated can be deeply analyzed to extract text feature information such as the wind power application scenario to which the document belongs, the distribution characteristics of professional terms, the text sentence structure, and the document type and format attributes. Combined with user-preset configuration parameters, the target translation requirements for this translation task can be determined, such as the translation language, translation accuracy standards, terminology standardization requirements, Chinese-English bilingual output requirements, and professional terminology highlighting requirements, etc.

[0053] S208. Retrieve the preset terminology resource library, match the terminology constraints corresponding to the text feature information and target translation requirements, and generate a terminology constraint set.

[0054] In this embodiment, a pre-built terminology resource library can be retrieved. Combined with the extracted document text feature information and the determined target translation requirements, precise matching and filtering are performed to select professional terms, standard translations, and prohibited translations that are suitable for the current document scenario and conform to translation standards. The various terminology constraint rules obtained from the filtering are then merged, integrated, and sorted to generate a terminology constraint set.

[0055] S209. Structure and bind the target translation unit and term constraint set to construct the translation task.

[0056] In this embodiment, the target translation unit and the term constraint set can be structurally encapsulated and associated to construct a translation task with a standardized format.

[0057] S210. Input the translation task into the preset translation model so that the preset translation model can combine the terminology constraint set in the translation task to perform translation on each target translation unit and generate the original translation corresponding to the translation task.

[0058] In this embodiment, the translation task can be input into a preset translation model, and the original translation corresponding to the translation task can be generated by the preset translation model.

[0059] S211. Match the protected object corresponding to each target placeholder according to the mapping relationship, and replace the corresponding target placeholder in the original translation with the matched protected object in turn to obtain the translation to be verified.

[0060] In this embodiment, the mapping relationship can be retrieved, all target placeholders contained in the original translation can be traversed, and the protection object corresponding to each target placeholder can be matched. All target placeholders are then sequentially restored to their corresponding protection objects as the translation to be verified.

[0061] S212. Perform consistency verification on the translation to be verified according to the term constraint set and the mapping relationship between the protected object and the target placeholder.

[0062] In this embodiment, the translation task can be validated using both the terminology constraint set and the pre-stored mapping relationships as dual verification benchmarks. This involves checking whether the translations of all wind power-related terms in the translation conform to standard constraints, whether the terminology throughout the text is consistent, whether there are any prohibited translations, and whether all protected objects are completely restored.

[0063] S213. Identify any abnormal information that occurs during the consistency verification process, generate the corresponding verification results, and determine the correction suggestions associated with the verification results.

[0064] Among them, the correction suggestions are standardized rectification suggestions automatically generated for various verification anomalies, which are used to guide the correction of anomaly information.

[0065] In this embodiment, anomalies occurring during the consistency verification process can be extracted, and all anomaly information, including anomaly location, anomaly terminology, erroneous translations, and defects in the restoration of protected objects, can be recorded. All anomaly information is then summarized and structured to form the verification results. Simultaneously, by combining preset engineering document translation standards and other information, correction suggestions can be automatically generated for various anomaly types.

[0066] S214. Identify the business terminology in the translation to be verified and highlight the business terminology.

[0067] Among them, business terminology specifically refers to industry-specific expressions such as professional terms, phrases, process descriptions, equipment names, and parameter terms that are unique to the engineering field.

[0068] In this embodiment, the business terminology contained in the translation to be verified can be accurately identified based on a domain-specific terminology database, and differentiated highlighting processing can be performed on the business terminology.

[0069] S215. Based on the backfilling positioning information, the translation to be verified is backfilled into the corresponding position of the document to be translated in the form of a comparison between the original text and the translation to be verified to generate the target translation. The target translation, verification results and correction suggestions are then output together.

[0070] In this embodiment, the backfilling location information corresponding to each translatable unit can be retrieved. The translation to be verified, which has completed consistency verification and terminology highlighting, is backfilled into the corresponding position in the document to be translated using a source-translation comparison layout to obtain the target translation. The generated target translation, complete verification results, and corresponding correction suggestions are then linked and output.

[0071] This invention identifies different layout structures within the document to be translated, determines the content to be translated according to the layout structure, and splits the content to be translated into at least two translatable units at the granular level of complete paragraphs and complete cells. It collects the position and sequence information of each translatable unit within the document to be translated, and generates corresponding backfill positioning information based on the position and sequence information, ensuring orderly and accurate translation and lossless backfilling of the document to be translated. By identifying preset identifier content contained in each translatable unit and using this preset identifier content as protected objects, it assigns target placeholders to each protected object, replaces the corresponding protected object in the translatable unit with the target placeholder, and generates the processed target translation unit. It records the correspondence between protected objects and corresponding target placeholders to form a mapping relationship, effectively solving the defects of general translation models that easily mistranslate, arbitrarily modify, or lose project numbers, parameter values, and standard identifiers, thus avoiding erroneous translation of the document to be translated. By extracting textual feature information from the text content of the document to be translated and determining the target translation requirements of the document, a pre-set terminology resource library is retrieved, and terminology constraints corresponding to the textual feature information and target translation requirements are matched to generate a terminology constraint set. The target translation units and the terminology constraint set are structured, encapsulated, and bound together to construct a translation task. The translation task is then input into a pre-set translation model, which combines the terminology constraint set within the translation task to perform translation on each target translation unit, generating the original translation corresponding to the translation task, thus achieving a complete translation of the content to be translated. By matching the protected object corresponding to each target placeholder according to the mapping relationship, the matched protected object replaces the corresponding target placeholder in the original translation in turn to obtain the translation to be verified. Consistency verification is performed on the translation to be verified according to the terminology constraint set and the mapping relationship between the protected object and the target placeholder. Anomalies that occur during the consistency verification process are identified, corresponding verification results are generated, and correction suggestions associated with the verification results are determined. Business professional terms in the translation to be verified are identified and highlighted. Based on the backfilling positioning information, the translation to be verified is backfilled into the corresponding position of the document to be translated in the form of comparison between the original text and the translation to be verified to generate the target translation. The target translation, verification results, and correction suggestions are linked and output to improve the practicality and traceability of the document. It completely solves the problems of paragraph disorder, table content misalignment, and loss of layout structure that are easy to occur in the batch processing of documents by traditional translation models. It preserves the original layout format and hierarchical structure of the engineering document to the greatest extent and greatly improves the format fidelity of the translation and the standardization of the engineering document.

[0072] Example 3 In one embodiment, this embodiment uses wind power engineering technical documents as the document to be translated, with the document format being .docx (OOXML). It uses a document translation system to execute a document translation method, and placeholders as target placeholders, to further illustrate a document translation method. This solution aims to address the technical pain points in wind power technical document translation, such as inconsistent terminology, low translation efficiency, and poor timeliness of the terminology database. Through the collaborative linkage of various units, it possesses high adaptability, high accuracy, high timeliness, and scalability, effectively meeting the core requirements of professionalism and efficiency in wind power translation. The document translation system includes a document intelligent parsing unit, an intelligent translation unit, an automated iteration unit for the wind power terminology database, and an automated translation document generation unit. Its application process is as follows: The document intelligent parsing unit is used to identify the structure and break down the content of input wind power engineering technical documents, recognizing structural information such as the cover, approval page, revision page, table of contents, and main text. Based on business requirements and standard document specifications, the table of contents and main text need to be translated, while images and formulas are not translated. After receiving the document uploaded by the user, the system will protect the identified cover, approval page, revision page, images, and formulas; simultaneously, it will break down the main text into the smallest translatable units according to the original format of wind power documents and record the sentence order after splitting to ensure granular alignment and meet the requirements of wind power business expression. The specific process is as follows: Firstly based on Tools such as [tool name] read the format as The wind power engineering technical document (OOXML) is constructed in memory using object mappings of paragraphs, tables, and cells. During parsing, the entire document is not converted to JavaScript Object Notation (JSON) or HyperText Markup Language (HTML). Instead, it is traversed step-by-step according to business rules. Tools such as `detect_cover_region` automatically detect and locate boundaries of regions in the Word document, and keyword recognition is used to locate the boundaries between the cover and body text. Cover pages, image paragraphs, tables of contents, and revision pages are skipped or protected. Existing reference content (historical translations) is cleaned up, retaining only the Chinese text to be translated. For each location to be translated, its plain text string is extracted and recorded in a data structure or object name structure, including paragraph / cell references and type tags (such as regular paragraphs, headings, numbered headings, table of contents lines, and table cells). Thus, the document is divided into a list of text to be translated (content to be translated) in the same order as the original text. The list is divided at the granularity of whole paragraphs or whole cells, without further splitting by sentence.

[0073] The intelligent translation unit performs controlled translation of translatable content in wind power engineering technical documents after the document intelligent parsing unit has completed structure recognition, content segmentation, and protection object isolation. Instead of directly translating the entire document, the intelligent translation unit uses the translatable units formed after document parsing as the smallest processing object. It employs a controlled translation algorithm based on structured translation task packages to complete translatable unit indexing, protection object locking, terminology constraint generation, translation generation, post-translation verification, anomaly marking, and correction suggestion output. The specific process is as follows: Step 1: Create a translatable unit structure index.

[0074] The system receives a set of translatable units output by the document intelligent parsing unit. Each translatable unit is represented as a structured data object, including at least a unit identifier, original text content, chapter / section, content type, adjacent units, original layout position, associated protected objects, and backfill identifier (backfill positioning information). Content types include chapter titles, body paragraphs, table text, figure captions, and parameter descriptions. Based on the unit identifier and original layout position, the system establishes a correspondence between the translatable unit and the original document page position, contextual content, and subsequent backfill position to ensure that the source of the original text can be accurately traced and backfilled to the corresponding position after the translation is generated.

[0075] Step 2: Identify and lock the protected object.

[0076] The system identifies equipment models, drawing numbers, standard numbers, formulas, numerical values, units of measurement, table fields, image labels, and other content that should not be translated or modified within translatable units, generating a list of protected objects. For each protected object, the system uses placeholder replacement to lock it and records the mapping relationship between the placeholder and the original protected object. For example, equipment models, standard numbers, formulas, or unit expressions are replaced with corresponding protected placeholders, preventing the preset translation model from directly processing such content when generating the translation. After the translation is generated, the system restores the protected object to its corresponding position in the translation based on the placeholder mapping relationship, thus avoiding problems such as mistranslation of equipment models, modification of numbers, destruction of formulas, or loss of units.

[0077] Step 3: Generate a set of term constraints.

[0078] The system generates a set of terminology constraints corresponding to the current translation task based on the original text, chapter, document type, content type, and context of the translatable unit. This set includes at least the original term, standard translation, applicable scenarios, contextual explanations, prohibited translations, term source, and term version information. When multiple candidate translations exist for the same term, the system does not determine the final translation solely based on word frequency or common translations. Instead, it combines the chapter, document type, contextual semantics, applicable scenarios, and term version priority of the translatable unit to determine the preferred translation for this task and adds it to the terminology constraint set. This set of terminology constraints only constrains the current translation generation process and does not directly trigger terminology database updates.

[0079] Step 4: Construct a structured translation task package.

[0080] The system combines the original text of translatable units, chapter context, adjacent units, protected object placeholders, terminology constraint sets, numbering and unit retention rules, prohibited translation rules, output format requirements, and wind power engineering document translation specifications into a structured translation task package. The structured translation task package includes at least the target language, the original text to be translated, a context window, required standard terminology, prohibited translations, placeholders to be preserved, numbering and unit retention rules, and translated text output format requirements. Through this structured translation task, the system transforms ordinary text translation tasks into translation tasks with terminology constraints, context constraints, structural constraints, and protected object constraints.

[0081] Step 5: Generate controlled translation (translation to be verified).

[0082] The system inputs a structured translation task package into the translation model, which then generates a translation under constraints of terminology, context, protected object, and output format. For translatable units containing multiple wind power technical terms, the translation should use the standard translation defined in the terminology constraint set. For translatable units containing equipment models, drawing numbers, standard numbers, formulas, numerical values, and units of measurement, the translation should retain the integrity of the corresponding placeholders and should not be translated, deleted, or rewritten. After the translation model outputs the translation, the system restores the protected object to the corresponding position in the translation based on the placeholder mapping relationship, forming the translation to be verified.

[0083] Step 6: Perform post-translation constraint consistency check.

[0084] The system automatically verifies the translation to be verified based on the terminology constraint set, the list of protected objects, the placeholder mapping relationship, and preset verification rules. The verification content includes at least terminology consistency verification, prohibited translation verification, placeholder integrity verification, number retention verification, unit retention verification, formula retention verification, translation missing verification, and contextual consistency verification. Specifically, terminology consistency verification determines whether the translation uses the standard translation from the terminology constraint set; prohibited translation verification determines whether the translation contains unrecommended or prohibited translations; placeholder integrity verification determines whether the protected objects in the original text are fully retained in the translation; and number retention, unit retention, and formula retention verification determine whether equipment models, drawing numbers, standard numbers, units of measurement, and formulas have been mistranslated, deleted, or rewritten, respectively.

[0085] Step 7: Generate anomaly markers and correction suggestions.

[0086] When the system detects issues such as the use of non-standard terminology, inconsistent translations of the same term, the appearance of prohibited translations, missing protected objects, rewritten equipment numbers, abnormal unit expressions, corrupted formulas, or missing translation content, it generates corresponding anomaly markers. These anomaly markers are linked to the translatable unit identifier, the original text location, the original text content, the translated fragment, the terminology constraint set, the protected object mapping relationship, and the triggered verification rules. For issues that can be directly located based on the terminology constraint set or placeholder mapping relationship, the system generates correction suggestions, such as replacing incorrect translations with standard translations, restoring rewritten equipment models, or adding missing units or formulas. For issues that cannot be directly determined by rules, the system does not directly modify the official translation but generates review prompts for further confirmation by reviewers.

[0087] Step 8: Output the backfillable translation and verification results.

[0088] The system binds validated translations, failed validation anomaly markers, correction suggestions, and review prompts to their corresponding translatable unit identifiers, creating backfillable translation results. These backfillable results are used for subsequent automated generation of translation document units to backfill translations based on their original layout positions. Simultaneously, the system maintains the relationships between translatable units, terminology constraint sets, protected object mappings, post-translation validation results, and correction records, providing reviewers with traceability evidence regarding translation sources, terminological basis, reasons for anomalies, and the modification process.

[0089] Through the above steps, the intelligent translation unit transforms the translation process of wind power engineering technical documents from ordinary text translation into a controlled translation process based on translatable unit structure indexing, protected object placeholder mapping, terminology constraint sets, and post-translation constraint consistency verification. This process reduces the risks of mistranslation of wind power terminology, multiple translations of the same term, incorrect modification of equipment numbers, loss of formulas or units, and errors in translation backfilling, thereby improving the consistency, accuracy, proofreadability, and traceability of translated wind power engineering technical documents.

[0090] The automated iteration unit for the wind power terminology database is used for the continuous maintenance and controlled updates of the wind power terminology resource database. It addresses issues such as scattered terminology sources, lack of basis for terminology updates, difficulty in tracing manual maintenance, and the potential impact of directly adding new terms to the database on translation consistency during the translation of wind power engineering technical documents. Instead of automatically adding newly identified terms or candidate translations to the official terminology database, this unit generates candidate terminology update suggestions based on terminology usage records during the translation process, post-translation verification anomalies, manual review and modification records, and historical document reuse. Controlled iteration of the terminology resource database is then achieved through manual review and version control mechanisms.

[0091] The specific process is as follows: Step 1: Collect terminology usage and review feedback data.

[0092] During the translation, post-translation verification, and manual review processes performed by the intelligent translation unit, the system records the terminology hit rate, missing words, standard translation usage, disabled translation method triggers, post-translation verification anomalies, manual modifications, and review comments for each translatable unit. These records are associated with the corresponding document structure unit, source text location, translated text segment, chapter, document type, and terminology constraint set, forming a terminology usage log and a review feedback log.

[0093] Step 2: Generate a set of candidate term sources.

[0094] Based on terminology usage logs and review feedback logs, the system extracts candidate terminology sources that may require maintenance. These candidate terminology sources include at least: wind power terminology expressions that appeared multiple times during the translation process but were not found in the terminology database; words that were repeatedly modified to the same translation by human reviewers; words that triggered terminology inconsistency anomalies multiple times during post-translation verification; and equipment names, component names, process actions, parameter expressions, and standard numbers that appear repeatedly in similar wind power engineering technical documents but lack standard translations.

[0095] Step 3: Generate candidate term update suggestions.

[0096] The system merges and deduplicates the sources of candidate terms, generating suggested updates. These updates include at least the original term, suggested translation, candidate type, frequency of occurrence, location of occurrence, contextual examples, associated document type, relevant translatable unit identifiers, manual modification records, historical translations, reasons for recommendation, and confidence markers. When multiple translations exist for the same term, the system does not directly determine the final translation but retains multiple candidate translations and their corresponding contexts for subsequent reviewers to assess.

[0097] Step 4: Perform manual review and terminology confirmation.

[0098] The system pushes suggested updates to candidate terms to the terminology review interface, where reviewers or terminology administrators confirm, modify, merge, reject, or mark as pending. Reviewers can determine whether a candidate term should be included in the official terminology resource library based on its location, contextual examples, historical translations, manual modification records, and associated document types. For approved candidate terms, the system generates an official terminology record; for rejected candidate terms, the system retains the reasons for rejection to prevent repeated recommendations.

[0099] Step 5: Create a terminology version record.

[0100] For approved terms, the system writes them into the wind power terminology resource library and generates corresponding terminology version records. Each terminology version record includes at least the original term, standard translation, applicable scenarios, contextual explanations, prohibited translations, terminology source, reviewer, review date, version number, scope of application, and historical change information. When existing terms undergo translation adjustments, changes in applicable scenarios, or updates to prohibited translations, the system does not directly overwrite the original record but generates a new terminology version while retaining historical versions to trace the terminology used in different stages of translation tasks.

[0101] Step 6: Update the terminology constraint generation rules.

[0102] Once the terminology version record is established, the system updates the terminology constraint generation rules in subsequent intelligent translation units based on the terminology's applicable scenarios, document types, chapter scope, and contextual descriptions. When generating terminology constraint sets, subsequent translation tasks select applicable terms based on the current terminology version, document type, and the context of the translatable unit, avoiding inconsistencies in translations caused by mixing different version terms.

[0103] Step 7: Track the effectiveness of terminology application.

[0104] During subsequent translation and review processes, the system continuously records the hit rate of new version terms, the adoption rate of translations, the number of manual modifications, and the occurrence of anomalies. This information is used to determine whether the terminology version is suitable for the actual wind power engineering technical document translation scenario. When a term continuously triggers manual modifications or post-translation verification anomalies during subsequent use, the system re-includes it in the candidate terminology update suggestions, entering the next round of review and version maintenance process.

[0105] Through the above process, this unit transforms the terminology database iteration process from a simple automatic update into a controlled maintenance process based on translation logs, review records, anomaly checks, and version management. This process avoids unapproved candidate terms directly affecting the final translation, while ensuring that the source, reason for modification, scope of application, and historical versions of each term are traceable, thereby improving the stability, accuracy, and maintainability of the wind power terminology database.

[0106] The automated translation document generation unit is used to automatically backfill the original Word wind power engineering technical document with Chinese and English side-by-side information after translation is completed, according to the split order and original format information, generating a translated document that maintains the original structure and format. Specialized terms in the wind power field are automatically highlighted, making it easier for reviewers to quickly assess the accuracy of the translation. Protected content such as the cover, approval page, revision pages, images, and formulas are retained as is without modification. The specific process is as follows: A self-built translation interface is called via requests (a third-party HTTP library), sending the list of texts to be translated in sequence (POST request) (JSON format). The returned translations and terminology details are parsed to obtain the sequentially aligned translation results. During backfilling, a new document is not created; instead, the original Document object is manipulated, and each translation is written back to the corresponding paragraph or cell using zip traversal.

[0107] The core processing logic resides in the translation backfilling for paragraphs (e.g., using `process_paragraph_with_translation`) and table cells (e.g., using `process_cell_with_translation`): First, the Style Preservation Manager and Font Unification Manager save style snapshots before clearing the content. Then, the styles are reconstructed according to the original text, line breaks, and translation, with optional terminology highlighting. Simultaneously, edge cases such as first-line indentation merging, second-level heading numbering, skipping table of contents rows, and compressing blank spaces within tables are handled. For numerical headings with missing numbers returned by the Application Programming Interface (API), the original heading numbers are automatically reconstructed into the translation.

[0108] Before saving, the table of contents and heading styles are modified at the OOXML level, and the final output is a complete .docx file. The overall process is a sequential alignment translation method based on in-situ rewriting and style snapshot restoration.

[0109] Example 4 Figure 3 This is a schematic diagram of the structure of a document translation device according to Embodiment 4 of the present invention. Figure 3 As shown, the device includes: a content extraction module 31, a relationship determination module 32, a task construction module 33, an initial translation module 34, and a translation generation module 35.

[0110] The content extraction module 31 is used to perform document parsing on the document to be translated, determine the content to be translated in the document, split the content to be translated into at least two translatable units, and determine the backfilling location information of each translatable unit.

[0111] The relationship determination module 32 is used to determine the protected object in each translatable unit, obtain the target translation unit by replacing the protected object with the target placeholder, and establish a mapping relationship between the protected object and the target placeholder.

[0112] The task construction module 33 is used to generate a set of terminology constraints based on the text content of the document to be translated and the target translation requirements, integrate the target translation unit and the set of terminology constraints, and construct the translation task.

[0113] The initial translation module 34 is used to input the translation task into the preset translation model, generate the original translation corresponding to the translation task through the preset translation model, and restore the protected object to the original translation according to the mapping relationship to obtain the translation to be verified.

[0114] The translation generation module 35 is used to perform consistency checks on the translation to be checked and generate the check results, and generate the target translation based on the check results and backfilling location information.

[0115] The technical solution of this invention involves a content extraction module that performs document parsing on the document to be translated, determining the content to be translated, splitting the content into at least two translatable units, and determining the backfilling location information for each translatable unit. A relationship determination module determines the protected objects in each translatable unit, replaces the protected objects with target placeholders to obtain the target translation unit, and establishes a mapping relationship between the protected objects and the target placeholders. A task construction module generates a set of terminology constraints based on the text content of the document to be translated and the target translation requirements, integrates the target translation units and the set of terminology constraints, and constructs a translation task. An initial translation module inputs the translation task into a preset translation model, generates the original translation corresponding to the translation task through the preset translation model, restores the protected objects to the original translation according to the mapping relationship to obtain the translation to be verified, and a translation generation module performs consistency verification on the translation to be verified and generates a verification result. Based on the verification result and backfilling location information, a target translation is generated. This reduces the risk of mistranslation of professional terms, multiple translations of the same term, and translation backfilling errors, and improves the accuracy, proofreadability, and traceability of engineering and technical document translations.

[0116] In one embodiment, the content extraction module 31 includes: The content determination unit is used to identify different layout structures within the document to be translated and to determine the content to be translated according to the layout structure. Content splitting unit is used to split the content to be translated into at least two translatable units, with the granularity of complete paragraphs and complete cells. The information extraction unit is used to collect the location and sequence information of each translatable unit in the document to be translated, and generate corresponding backfill positioning information according to the location and sequence information.

[0117] In one embodiment, the relationship determination module 32 includes: The object extraction unit is used to identify the preset identifier content contained in each translatable unit and to treat the preset identifier content as the protected object. The placeholder allocation unit is used to assign target placeholders to each protected object one by one, and replace the corresponding protected object in the translatable unit with the target placeholders to generate the processed target translation unit. The mapping unit is used to record the mapping relationship between the protected object and the corresponding target placeholder.

[0118] In one embodiment, the task construction module 33 includes: The feature extraction unit is used to extract text feature information according to the text content of the document to be translated and to determine the target translation requirements of the document to be translated. The set generation unit is used to retrieve a preset terminology resource library, match terminology constraints corresponding to text feature information and target translation requirements, and generate a terminology constraint set. The task construction unit is used to structurally encapsulate and associate target translation units and terminology constraint sets to construct translation tasks.

[0119] In one embodiment, the initial translation module 34 includes: The translation generation unit is used to input the translation task into the preset translation model, so that the preset translation model, in combination with the set of term constraints within the translation task, performs translation on each target translation unit and generates the original translation corresponding to the translation task. The translation replacement unit is used to match the protection object corresponding to each target placeholder according to the mapping relationship, and replace the corresponding target placeholder in the original translation with the matched protection object in turn to obtain the translation to be verified.

[0120] In one embodiment, the translation generation module includes: The verification unit is used to perform consistency verification on the translation to be verified according to the term constraint set and the mapping relationship between the protected object and the target placeholder; The result generation unit is used to identify the abnormal information that occurs during the consistency verification process, generate the corresponding verification results, and determine the correction suggestions associated with the verification results. The marking unit is used to identify business terminology in the translation to be verified and to highlight the business terminology. The translation output unit is used to backfill the translation to be verified into the corresponding position of the document to be translated based on the backfill positioning information, in the form of the original text and the translation to be verified being arranged in comparison, to generate the target translation, and to output the target translation, verification results and correction suggestions in association.

[0121] In one embodiment, the document translation apparatus further includes: The log generation module is used to collect the terminology constraints corresponding to the abnormal information, determine the modification records corresponding to the abnormal information, and generate the review feedback log by combining the terminology constraints and modification records. The suggestion generation module is used to extract candidate terms from the review feedback log and generate candidate term update suggestions that include the original text of the candidate term, the candidate translation, the frequency of occurrence, the context, the manual modification record, and the recommendation reason. The judgment module is used to push the candidate term update suggestions to the review end, and the review end determines whether the candidate term is the target term. The terminology update module is used to store the candidate term in a preset terminology resource library when the candidate term is determined to be the target term, and to update the generation rules of terminology constraints according to the applicable scenarios of the candidate term.

[0122] The document translation apparatus provided in this embodiment of the invention can execute the document translation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0123] Example 5 Figure 4 This is a schematic diagram of the structure of an electronic device implementing the document translation method of this invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0124] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0125] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0126] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as document translation methods.

[0127] In some embodiments, the document translation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the document translation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the document translation method by any other suitable means (e.g., by means of firmware).

[0128] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0129] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0130] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0131] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0132] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0133] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0134] In one embodiment, the present invention further includes a computer program product, which includes a computer program that, when executed by a processor, implements the document translation method of any embodiment of the present invention.

[0135] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0137] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A document translation method, characterized in that, include: The document to be translated is parsed to determine the content to be translated in the document, the content to be translated is split into at least two translatable units, and the backfilling positioning information of each translatable unit is determined. The protected object in each of the translatable units is determined, and the target translation unit is obtained by replacing the protected object with the target placeholder. A mapping relationship between the protected object and the target placeholder is established. A set of terminology constraints is generated based on the text content of the document to be translated and the target translation requirements. The target translation unit and the set of terminology constraints are then integrated to construct a translation task. The translation task is input into a preset translation model, and the original translation corresponding to the translation task is generated by the preset translation model. The protected object is restored to the original translation according to the mapping relationship to obtain the translation to be verified. Perform a consistency check on the translation to be checked and generate a check result. Generate the target translation based on the check result and the backfilling positioning information.

2. The method according to claim 1, characterized in that, The document to be translated is parsed to determine the content to be translated in the document, the content to be translated is split into at least two translatable units, and the backfilling location information of each translatable unit is determined, including: Identify different layout structures within the document to be translated, and determine the content to be translated according to the layout structure; The content to be translated is split into at least two translatable units, using complete paragraphs and complete cells as the granularity. The position and sequence information of each translatable unit in the document to be translated are collected, and corresponding backfill positioning information is generated according to the position and sequence information.

3. The method according to claim 1, characterized in that, The step of determining the protected object in each of the translatable units, obtaining the target translation unit by replacing the protected object with a target placeholder, and establishing a mapping relationship between the protected object and the target placeholder includes: Identify the preset identifier content contained in each of the translatable units, and use the preset identifier content as the protected object; For each of the protected objects, a target placeholder is assigned, and the corresponding protected object in the translatable unit is replaced with the target placeholder to generate the processed target translation unit; The mapping relationship between the protected object and the corresponding target placeholder is recorded.

4. The method according to claim 1, characterized in that, The process of generating a terminology constraint set based on the text content of the document to be translated and the target translation requirements, integrating the target translation unit and the terminology constraint set, and constructing a translation task includes: Extract text feature information from the text content of the document to be translated, and determine the target translation requirements of the document to be translated; Retrieve a preset terminology resource library, match the terminology constraints corresponding to the text feature information and the target translation requirements, and generate a terminology constraint set; The target translation unit and the term constraint set are structured, encapsulated, and associated to construct a translation task.

5. The method according to claim 1, characterized in that, The process of inputting the translation task into a preset translation model, generating the original translation corresponding to the translation task through the preset translation model, and restoring the protected object to the original translation according to the mapping relationship to obtain the translation to be verified includes: The translation task is input into a preset translation model, so that the preset translation model combines the terminology constraint set within the translation task to perform translation on each of the target translation units, generating the original translation text corresponding to the translation task; Based on the mapping relationship, the protection object corresponding to each target placeholder is matched, and the matched protection object is used to replace the corresponding target placeholder in the original translation in turn to obtain the translation to be verified.

6. The method according to claim 1, characterized in that, The process of performing a consistency check on the translation to be checked and generating a check result, and generating a target translation based on the check result and the backfilling positioning information, includes: Perform a consistency check on the translation to be verified according to the term constraint set and the mapping relationship between the protected object and the target placeholder; Identify any anomalies that occur during the consistency verification process, generate the corresponding verification results, and determine any corrective suggestions associated with the verification results. Identify the business terminology in the translation to be verified and highlight the business terminology. Based on the backfilling positioning information, the translation to be verified is backfilled into the corresponding position of the document to be translated in the form of a comparison between the original text and the translation to be verified to generate the target translation. The target translation, the verification result, and the correction suggestion are then output in association.

7. The method according to claim 6, characterized in that, After performing a consistency check on the translation to be checked and generating a check result, and generating the target translation based on the check result and the backfilling positioning information, the process further includes: Collect the terminology constraints corresponding to the anomaly information, determine the modification records corresponding to the anomaly information, and generate a review feedback log by combining the terminology constraints and the modification records. Candidate terms are extracted from the review feedback log, and update suggestions for candidate terms are generated, which include the original text of the candidate terms, the candidate translations, the frequency of occurrence, the context, the records of manual modifications, and the reasons for recommendation. The updated candidate terms are pushed to the review panel, which then determines whether the candidate terms are target terms. When a candidate term is determined to be a target term, the candidate term is stored in a preset term resource library, and the generation rules of term constraints are updated according to the applicable scenarios of the candidate term.

8. A document translation device, characterized in that, include: The content extraction module is used to perform document parsing on the document to be translated, determine the content to be translated in the document to be translated, split the content to be translated into at least two translatable units, and determine the backfilling positioning information of each translatable unit; The relationship determination module is used to determine the protected object in each of the translatable units, obtain the target translation unit by replacing the protected object with the target placeholder, and establish a mapping relationship between the protected object and the target placeholder; The task construction module is used to generate a set of terminology constraints based on the text content of the document to be translated and the target translation requirements, and to integrate the target translation unit and the set of terminology constraints to construct a translation task; An initial translation module is used to input the translation task into a preset translation model, generate the original translation corresponding to the translation task through the preset translation model, and restore the protected object to the original translation according to the mapping relationship to obtain the translation to be verified. The translation generation module is used to perform consistency verification on the translation to be verified and generate verification results, and generate target translation based on the verification results and the backfill positioning information.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the document translation method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the document translation method according to any one of claims 1-7.

Images