Software internationalization processing method, computing device and readable storage medium
By combining a shared thesaurus and translation model with a review process, the problem of fragmented and scattered translation resources in software internationalization has been solved, achieving efficient and accurate translation management and improving translation efficiency and the usability of the shared thesaurus.
Patent Information
- Application Number
- CN202511392782.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-26
- Publication Date
- 2025-12-23
Smart Images
Figure CN121189345A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a software internationalization processing method, a computing device, and a computer-readable storage medium. Background Technology
[0002] With the acceleration of globalization, software products need to be geared towards users in different language environments, making software internationalization a crucial aspect of software development. Software internationalization refers to designing software products into universal versions that can adapt to different language and cultural environments during the software design and development process. This facilitates subsequent localization efforts, meaning translating and adapting the software interface, documentation, and other content to the local language and cultural habits based on the needs of users in specific regions or languages.
[0003] In existing technologies, when software products achieve internationalization, they usually need to manually input and manage the translation of terms, resulting in problems such as scattered translation resources, fragmented management, and high barriers to reuse. Summary of the Invention
[0004] The purpose of this application is to provide a software internationalization processing method, computing device, and computer-readable storage medium that can solve the problems of scattered translation resources, fragmented management, and high reuse barriers in the prior art through a shared term library shared by multiple business modules, and can also ensure the high availability of the shared term library.
[0005] To achieve the above objectives: In a first aspect, embodiments of this application provide a software internationalization processing method, including the following steps: S11: obtaining terms to be translated from a shared term library, wherein the translation results in the shared term library are used by multiple business modules; S12: calling a translation model to translate the terms to be translated, obtaining the corresponding first translation result, and adding the first translation result to the shared term library; S13: entering a review process based on the terms to be translated or the first translation result to obtain the corresponding second translation result, and synchronizing the second translation result to the shared term library.
[0006] In one embodiment, the proofreading process includes: adding the term to be translated or the first translation result to the proofreading library; creating one or more proofreading tasks based on the proofreading library, wherein the proofreading tasks include a preset number of terms to be translated or the first translation result; issuing the proofreading tasks to send corresponding task notifications to one or more proofreading terminals; receiving at least one second translation result after the proofreading terminal accepts and completes the proofreading task, wherein the proofreading terminal completes the proofreading task by means of manual proofreading and / or model proofreading; and / or recording the task completion data of the proofreading terminal, and / or executing incentive control corresponding to the proofreading terminal according to preset rules and task completion data.
[0007] In one embodiment, adding the term to be translated or the first translation result to the review database includes: performing a duplicate detection on the review database based on the term to be translated or the first translation result; adding the term to be translated or the first translation result to the review database when the duplicate detection result indicates no duplication; or, performing deduplication processing on the review database when the duplicate detection result indicates duplication.
[0008] In one embodiment, after step S13, the method includes: triggering incremental training of the translation model based on the second translation result to fine-tune the translation model.
[0009] In one embodiment, step S12 includes: obtaining language requirements; obtaining related terms in a shared terminology library that have a contextual relationship with the term to be translated; and calling a translation model to perform translation based on the term to be translated, related terms, and / or language requirements to obtain a first translation result for the term to be translated, wherein the language requirements correspond to one or more languages, and the translation model is a generative model using AIGC technology.
[0010] In one embodiment, the method of this application further includes: receiving term resources imported by a business module, and adding at least a portion of the data in the term resources to a shared term library, wherein the at least a portion of the data includes the translation results of terms to be translated and / or terms that have already been translated.
[0011] In one embodiment, the method of this application further includes: responding to a translation acquisition request from a business module, when the translation acquisition request is the first request within a preset time limit, acquiring the corresponding translation result from a shared terminology library and sending it to the business module, and storing the translation result corresponding to the translation acquisition request in a Redis cache according to a preset time limit; or, when the translation acquisition request is a non-first request within a preset time limit, acquiring the corresponding translation result from a Redis cache and sending it to the business module; and / or, backing up some or all of the translation results in the shared terminology library to the business module according to a preset backup rule.
[0012] In one embodiment, the method of building the translation model includes: training the model framework using general corpora corresponding to multiple languages and / or terminology databases corresponding to multiple preset domains to obtain an initial translation model; and / or training the translation model using the translation results of translated terms in the terminology resources imported by the business module to update the translation model.
[0013] Secondly, embodiments of this application provide a computing device, including: a processor and a memory storing a computer program, wherein when the processor runs the computer program, the steps of the above-described software internationalization processing method are implemented.
[0014] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described software internationalization processing method.
[0015] The software internationalization processing method, computing device, and computer-readable storage medium provided in this application include the following steps: S11: Obtaining terms to be translated from a shared terminology library, wherein the translation results in the shared terminology library are used by multiple business modules; S12: Calling a translation model to translate the terms to be translated, obtaining the corresponding first translation result, and adding the first translation result to the shared terminology library; S13: Entering a review process based on the terms to be translated or the first translation result to obtain the corresponding second translation result, and synchronizing the second translation result to the shared terminology library. Thus, in the technical solution of this application, when a shared terminology library shared by multiple business modules contains terms to be translated, the terms to be translated can be translated using a translation model to quickly update the shared terminology library for use by each business module, ensuring the basic or emergency needs of each business module. Subsequently, a review process can be used to translate and review the model translation results or the aforementioned terms to be translated to obtain more accurate translation results, and the shared terminology library can be updated again for use by each business module, ensuring that each business module can obtain more accurate translation resources from the updated shared terminology library. Therefore, the technical solution of this application can solve the problems of scattered translation resources, fragmented management, and high reuse barriers in the existing technology by sharing a shared term library among multiple business modules, and can also ensure the high availability of the shared term library. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating the software internationalization processing method provided in an embodiment of this application.
[0017] Figure 2 This is a schematic diagram of the review process provided in the embodiments of this application.
[0018] Figure 3This is a schematic diagram of the workflow of the internationalization development management platform used as an example in this application.
[0019] Figure 4 A schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation
[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements.
[0021] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.
[0022] It should be understood that although the terms first, second, third, etc., may be used herein to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this document, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if," as used herein, can be interpreted as "when," "when," or "in response to determination." Furthermore, as used herein, the singular forms "a," "an," and "the" are intended to also include the plural forms unless the context indicates otherwise. It should be further understood that the terms "comprising," "including," indicate the presence of the stated feature, step, operation, element, component, item, kind, and / or group, but do not exclude the presence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, kinds, and / or groups. The terms "or" and "and / or" as used herein are to be interpreted as inclusive, or mean any one or any combination thereof. Therefore, "A, B, or C" or "A, B, and / or C" means "any one of the following: A; B; C; A and B; A and C; B and C; A, B, and C". Exceptions to this definition will only occur if the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.
[0023] It should be understood that although the steps in the flowcharts of this application's embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.
[0024] It should be noted that step designations such as S11 and S12 are used in this document for the purpose of more clearly and concisely describing the corresponding content, and do not constitute a substantial limitation on the order. In specific implementation, those skilled in the art may execute S12 first and then S11, etc., but these should all be within the protection scope of this application.
[0025] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.
[0026] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.
[0027] To facilitate understanding of this application, the following explanations are provided for the terms and technical objects that may be used in this application: Crowdsourcing can be characterized as an organizational model that publicly or selectively distributes specific tasks (especially those requiring human intelligence or judgment) to a large, undefined group of people (usually through online platforms) to complete them. A terminology database can represent a list or database containing key terms in a specific field, project, or product and their standardized, accepted translations in the target language; it defines a one-to-one mandatory or recommended correspondence between source language terms and target language terms. AIGC technology can be characterized by using artificial intelligence to automatically generate content such as text, images, and audio. Transformer-XL Architecture: Transformer-XL is an improved version of the Transformer architecture. By introducing mechanisms such as segment-level recurrence and relative positional encoding, it solves the limitations of the traditional Transformer in processing long texts. It can better capture the dependencies in long texts and is suitable for generating long and coherent content.
[0028] See Figure 1 This application provides a software internationalization processing method, which can be executed by a computing device provided in this application. The computing device can be implemented in software and / or hardware. In this embodiment, the executing entity of the method is a management platform deployed in a server. The software internationalization processing method provided in this embodiment includes steps (e.g., steps S11 to S13): S11: Obtain the terms to be translated from the shared term library, where the translation results in the shared term library are used by multiple business modules.
[0029] S12: Call the translation model to translate the term to be translated, obtain the corresponding first translation result, and add the first translation result to the shared term library.
[0030] S13: Based on the term to be translated or the first translation result, proceed to the proofreading process to obtain the corresponding second translation result, and synchronize the second translation result to the shared term library.
[0031] In one implementation, the shared terminology library can be used by various business modules to call translation results or export all or part of the translated resources in the shared terminology library according to their own needs.
[0032] In one embodiment, the terms to be translated in the shared term library can be new terms newly added to the shared term library, or terms that have been translated and marked as terms to be translated when a correction operation is triggered for a specific translation result in the shared term library.
[0033] In one embodiment, the translation result can be represented as any file format of translated content in at least one language associated with a specific term.
[0034] It should be understood that the first translation result is the translation result obtained by the translation model (or machine translation result), and the second translation result is the translation result obtained after entering the proofreading process, or secondary translation.
[0035] In one implementation, multiple business modules can be multiple independent business software or business systems, or multiple functional modules within a single software.
[0036] Through the above-described technical solution of this embodiment, when multiple business modules share a common terminology library containing terms to be translated, the library can be quickly and initially updated using a translation model to provide a preliminary update for each module. This ensures the basic or emergency needs of each business module are met. Subsequently, a review process can be used to verify the model translation results or the aforementioned terms to be translated, resulting in a more accurate translation. The shared terminology library is then updated again for each business module to ensure they can obtain more precise translation resources from the updated library. Therefore, the technical solution of this embodiment addresses the problems of fragmented translation resources, fragmented management, and high reuse barriers in existing technologies by using a shared terminology library shared by multiple business modules, while also ensuring the high availability of the shared terminology library.
[0037] In one embodiment, the proofreading process can characterize the specific steps of performing manual proofreading and / or model proofreading based on the translation results output by the translation model, or it can characterize the specific steps of performing manual secondary translation or other translation model secondary translation based on the terms to be translated input into the translation model, and comparing the two translation results.
[0038] In one embodiment, see Figure 2 The proofreading process includes, but is not limited to: S21: Add the term to be translated or the first translation result to the proofreading database; S22: Create one or more proofreading tasks based on the proofreading database. Each proofreading task includes a preset number of terms to be translated or the first translation result. S23: Issue the proofreading task to send the corresponding task notification to one or more proofreading clients; S24: Receive at least one second translation result from the proofreading end after accepting and completing the proofreading task, wherein the proofreading end completes the proofreading task through manual proofreading and / or model proofreading; and / or, S25: Record the task completion data of the reviewer, and / or execute incentive control corresponding to the reviewer based on preset rules and task completion data.
[0039] In one embodiment, the proofreading terminal can be a terminal or client used by proofreaders who perform manual proofreading, and / or it can be a terminal or functional module configured with other translation models.
[0040] For example, in the technical solution of this embodiment, after the management platform adds each first translation result to the proofreading database, it can periodically select first translation results from the proofreading database, create proofreading tasks according to preset creation rules (e.g., the number of first translation results for a single proofreading task, task completion requirements (e.g., task completion time, target language, etc.), and publish the proofreading tasks through a crowdsourcing mechanism to notify multiple proofreading ends to bid for the task or directly assign the task; after the proofreading personnel corresponding to the proofreading end that accepts the task complete the proofreading task and obtain the proofreading result (including at least one second translation result), the aforementioned proofreading end can... The proofreading results are fed back to the management platform, which then synchronizes at least one second translation result from the proofreading results to the shared terminology database (for example, if the first and second translation results for a specific term are inconsistent, the second translation result for that term replaces the first translation result in the shared terminology database). The management platform can use the task completion data of the proofreading end (e.g., the number of tasks completed within a preset time, the number of terms translated, etc.) to determine the performance or remuneration of the proofreaders. This embodiment's technical solution can build a crowdsourcing network of part-time translators through a shared architecture of crowdsourced proofreading and a terminology database, supporting task distribution and incentive settlement. Therefore, this embodiment's technical solution can improve the response speed of translation proofreading through a crowdsourcing mechanism, further enhancing the high availability of the shared terminology database.
[0041] In another implementation, the review process includes, but is not limited to: The term to be translated is translated using other translation models to obtain the corresponding reference translation results; When the first translation result and the reference translation result are consistent for the same term to be translated, the reference translation result shall be taken as the second translation result. When the first translation result and the reference translation result for the same term to be translated are inconsistent, the comparison information between the first translation result and the reference translation result will be added to the proofreading database. One or more review tasks are created based on the review database. Each review task includes a preset number of comparison information. The task of proofreading and reviewing is published to send the corresponding task notification to one or more proofreading and reviewing clients; The system receives at least one second translation result after the proofreading end has accepted and completed the proofreading task, wherein the proofreading end completes the proofreading task manually; and / or, Record the task completion data of the review and approval end, and / or execute incentive control corresponding to the review and approval end based on preset rules and task completion data.
[0042] Thus, in another implementation of this embodiment, when the first translation result matches the reference translation result generated by the second translation model, the reference translation result is directly synchronized to the shared terminology library as the second translation result, reducing the workload of subsequent manual review and improving review efficiency. When the first translation result does not match the reference translation result, the comparison information is added to the review library, and the review task is assigned to the review end for manual review by creating a review task. This mechanism ensures the accuracy of the translation result. In addition, by introducing at least two translation models to generate translation results and comparing them, the accuracy of the translation can be mutually verified. This multi-model verification mechanism can effectively reduce the errors that may be caused by a single translation model and improve the reliability of the translation result.
[0043] In one embodiment, the proofreading database can be associated with historical proofreading records, wherein the historical proofreading records include at least a unique identifier of the terminology to be translated or the first translation result that has flowed through the proofreading database and has completed proofreading and / or has been published after the proofreading task has been completed.
[0044] In one embodiment, adding the term to be translated or the first translation result to the review database includes: performing a duplicate detection on the review database based on the term to be translated or the first translation result; adding the term to be translated or the first translation result to the review database when the duplicate detection result indicates no duplication; or, performing deduplication processing on the review database when the duplicate detection result indicates duplication.
[0045] In one implementation, repeated detection may include at least one of the following: Check whether the term to be translated or the first translation result to be added to the current proofreading database already exists; Check if there are any identifiers in the historical proofreading records associated with the proofreading database that are to be added to the proofreading database or that correspond to the first translation result.
[0046] In one embodiment, the duplicate detection result represents the situation of non-duplication, including the absence of a term to be translated or a first translation result to be added to the current review database, and the absence of an identifier corresponding to a term to be translated or a first translation result to be added to the review database in the historical review records associated with the review database.
[0047] In one embodiment, the duplicate detection result characterizes the repetition, including any one of the following: The current proofreading database contains entries to be translated or first translation results that will be added to the database. The historical proofreading records associated with the proofreading database contain identifiers for terms to be translated or the first translation result that will be added to the proofreading database.
[0048] In one embodiment, the deduplication process may include at least one of the following: If there are entries to be translated or first translation results to be added to the current review database, or if there are identifiers corresponding to entries to be translated or first translation results to be added to the review database in the historical review records associated with the review database, then the duplicate addition of entries to be translated or first translation results to the review database will be prevented. If there are terms or first translation results in the current review database that are to be added to the review database, the task release priority of terms or first translation results corresponding to duplicate detection results in the review database will be increased so that they can be selected first when creating review tasks, thereby achieving priority review.
[0049] The above-described technical solution in this embodiment incorporates a proofreading and deduplication mechanism, which enables the same content (the term to be translated or the first translation result) to be proofread only once. Subsequently, it can be synchronized to all shared modules through a shared term library to eliminate duplicate work.
[0050] In one embodiment, step S13: after entering the proofreading process based on the term to be translated or the first translation result to obtain the corresponding second translation result, and synchronizing the second translation result to the shared term library, may include: triggering incremental training on the translation model based on the second translation result to fine-tune the translation model. Thus, the technical solution of this embodiment continuously improves the translation quality of the translation model by using the proofreading results to train the translation model in reverse.
[0051] In one embodiment, step S12: calling the translation model to translate the term to be translated, obtaining the corresponding first translation result, and adding the first translation result to the shared term library, may include: obtaining language requirements; obtaining related terms in the shared term library that have a contextual relationship with the term to be translated; and calling the translation model to translate according to the term to be translated, related terms, and / or language requirements to obtain the first translation result of the corresponding term to be translated, wherein the language requirements correspond to one or more languages, and the translation model is a generative model using AIGC technology.
[0052] Thus, in the technical solution of this embodiment, during the model translation process, related terms that have a contextual connection with the term to be translated can be introduced. This ensures that when the translation model processes the term to be translated, it can fully consider its usage and meaning in the actual use scenario or technical field, constrain the output of the translation model, and thus enable the translation model to generate more accurate and context-appropriate translation results.
[0053] Furthermore, in the technical solution of this embodiment, the translation model can be constrained by language requirements during the model translation process, so that the translation model can translate according to the characteristics and language habits of the target language corresponding to the language requirements. Since there are differences in grammar, vocabulary and cultural background between different languages, accurate matching of language requirements can ensure that the translation results output by the translation model conform to the expression habits and cultural background of the target language, thereby enabling the translation model to generate translation results that are more in line with the style of the target language.
[0054] In one embodiment, the method of building the translation model includes: training the model framework using general corpora corresponding to multiple languages and / or terminology databases corresponding to multiple preset domains to obtain an initial translation model; and / or training the translation model using the translation results of translated terms in the terminology resources imported by the business module to update the translation model.
[0055] In one implementation, the translation model's framework can be a variant combining the AIGC engine and the Transformer framework (e.g., the Transformer-XL architecture). This combination enables the translation model to more accurately understand the semantics of the source language text. The AIGC engine complements the Transformer framework's understanding capabilities through its generative abilities, allowing the model to better handle the ambiguity and diversity of language. Furthermore, the combination of the AIGC engine and the Transformer framework allows the translation model to adapt to multiple languages and domains. The AIGC engine learns from extensive multilingual data, mastering the expressions of different languages, while the Transformer framework can handle complex semantic structures across different domains.
[0056] In one embodiment, the general corpus includes multiple general corpus data, which represent the translation results of regular terms and their corresponding specific languages.
[0057] In the technical solution of this embodiment, when building the translation model, a general corpus set corresponding to multiple languages is used for training, so that the translation model can learn the basic language structure, grammatical rules and vocabulary usage of multiple languages. This multilingual training method enables the model to have a broad language knowledge base, so that it can more accurately understand and generate translation results when faced with translation tasks in different languages.
[0058] In addition, the terminology database contains key terms in specific fields and their standardized and recognized translations. These translations are usually verified by industry experts, which can ensure the professionalism and accuracy of the translation. By training the model framework through the terminology database, the subsequent translation model can directly use the accurate translations of these professional terms, avoiding errors caused by a lack of domain knowledge.
[0059] In this embodiment, the technical solution allows for the training of the translation model framework by combining general corpora corresponding to multiple languages and terminology databases corresponding to multiple preset domains. This enhances the multilingual adaptability and professionalism of the resulting translation model when performing translation tasks, and also improves the accuracy and consistency of the translation model.
[0060] In one implementation, the translation results of translated terms in the terminology resources imported by the business module are typically data generated during the actual development or use of the business module, or data maintained by personnel. Thus, by using the translation results of translated terms in the terminology resources imported by the business module to optimize and train the translation model, it can be ensured that the translation model can promptly reflect the latest language usage habits and actual business needs.
[0061] In one embodiment, the translation model may be configured with a conflict resolution mechanism, so that the translation model can generate a preliminary translation result based on the term to be translated and compare it with the standardized translation in the terminology database of the specific domain. When there is a difference between the preliminary translation result and the standardized translation in the terminology database of the specific domain, the conflict resolution mechanism is activated to determine the final output translation result.
[0062] In one implementation, the conflict resolution mechanism is based on a confidence-weighted voting method, for example, by weighting the voting based on the confidence level of the translation model and the priority of the terminology corpus. The confidence level of the translation model represents the confidence level assigned to each generated preliminary translation result, indicating the model's degree of confidence in the preliminary translation result; the priority of the terminology corpus indicates that terms and standardized translations within the corpus have different priorities, reflecting the importance and standardization of terms in a specific domain.
[0063] In this embodiment, the translation model, by introducing a conflict resolution mechanism, can comprehensively consider model confidence and terminology database priority, thus enabling more accurate translation of specialized terms. Therefore, the translation model in this embodiment ensures that the translation results both conform to the model's understanding of the context and closely align with the standardization requirements of the specific domain.
[0064] In one embodiment, the method provided in this embodiment may further include: receiving term resources imported by a business module, and adding at least a portion of the data from the term resources to a shared term library, wherein the at least a portion of the data includes terms to be translated and / or translation results of already translated terms. Thus, in this embodiment's technical solution, the management platform can grant import permissions to designated shared modules, enabling the shared modules to back up existing term resources to the management platform's shared term library to expand the content of the shared term library. Furthermore, the management platform can also import term resources from business modules to the shared term library via API calls to expand the content of the shared term library. Additionally, when a business module performs translation management itself (e.g., modifying the translation results of already translated terms in translation resources, or translating terms to be translated to obtain new translation results), the translated term resources can be sent to the shared term library for backup or shared with other business modules.
[0065] In one embodiment, the method provided in this embodiment may further include: responding to a translation retrieval request from a business module; when the translation retrieval request is the first request within a preset time limit, retrieving the corresponding translation result from a shared terminology library and sending it to the business module; and storing the translation result corresponding to the translation retrieval request in a Redis cache according to a preset time limit; or, when the translation retrieval request is not the first request within the preset time limit, retrieving the corresponding translation result from the Redis cache and sending it to the business module. Thus, the technical solution of this embodiment uses a Redis cache to store frequently used translation resources (e.g., translation results corresponding to non-first requests within a preset time limit) to improve the response efficiency of translation retrieval requests, and binds time-sensitive tags to frequently used resources to achieve automatic expiration and updates, ensuring that the Redis cache can better store commonly used translation resources.
[0066] In one implementation, after a business system obtains translation results through a Redis cache, if the business system modifies the translation results to obtain updated translation results, it can use the API interface configured on the management platform to reset the Redis cache. This allows the updated translation results obtained by the business modules to be reloaded individually or in batches according to a preset filtering range, so that each business system can immediately call the latest translation results as needed.
[0067] In one embodiment, the method provided in this embodiment may further include: backing up some or all of the translation results in the shared terminology library to the business module according to preset backup rules. For example, according to a preset backup frequency, all translation results of translated terms in the shared terminology library corresponding to a specific business module may be backed up to the specific business module.
[0068] Thus, the technical solution of this embodiment can provide business modules with at least two ways to obtain the translation results of translated terms in the shared terminology library: one is through on-demand invocation of translation retrieval requests, and the other is through exporting backups from the shared terminology library. Optionally, the business system can prioritize obtaining the translation results of translated terms in the shared terminology library through on-demand invocation of translation retrieval requests. If the invocation fails, the required translation results of translated terms can be obtained from the terminology resources backed up locally by the business module.
[0069] Based on the same inventive concept as the foregoing embodiments, the foregoing embodiments will be illustrated below by way of an international development management platform (i.e., the aforementioned management platform) that applies the above methods: I. Configuration of the international development management platform: (1) Computer: Used to run PC programs; (2) Relational database: used to store the identifier of the business system (i.e. the aforementioned business module), the operator permissions of the business system, the language of the business system, the terms, the translation of the terms into specific languages, the translation records of terms, the proofreading tasks, the qualifications of proofreaders and the proofreading settlement standards, etc. (3) Redis database; used to store the translation results corresponding to the terms, and can achieve microsecond-level response; (4) AIGC engine: used for multilingual translation and to support continuous optimization and iteration of the shared terminology database to improve translation results.
[0070] II. Workflow of the International Development Management Platform (see reference) Figure 3 ): 1. The international development management platform allows administrators to log in, enabling them to manage business systems through the platform. The management of business systems includes, but is not limited to: (1) Create business systems for different projects within an international development management platform; (2) In the international development management platform, operators with different permissions can be configured for each business system that has been created; (3) In the international development management platform, you can add or delete the language of the configuration of a specific business system.
[0071] 2. The international development management platform supports the input of terms and translation results; For example, the internationalization development management platform can grant input permissions to specific business systems, enabling these systems to import their own terminology resources (terms and / or translation results) into the internationalization development management platform. For example, the internationalization development management platform can support external systems or devices to make calls in the form of API interfaces, so that external systems or devices can import term resources through API interfaces; For example, after receiving the entered term resources (terms and / or translation results), the internationalization development management platform can import the term resources into a table according to specific import rules so that they can be stored in the shared term library later; for example, the terms in the imported term resources can be stored in the term table, and the translation results in the imported term resources can be stored in the translation text table. For example, the dimensions of the terminology table include, but are not limited to: ENTRY_ID (the primary key of the terminology table, which can be used to establish a connection with the translation text table), business system ID, encoding, language, text, type, category, translation completion status identifier, and internationalization key. The business system ID plus the internationalization key can uniquely identify a term. Encoding, type, and category support the differentiation of terms, facilitating batch processing by users. Furthermore, the terminology table can reserve multiple dimensions of fields (e.g., five dimensions) to support flexible expansion of new functions by various business systems, support recording original terms in text, and support recording the language of the original terms. For newly stored terminology data, the internationalization development management platform can automatically retrieve existing translation results based on the entered terminology resources and / or its own shared terminology library (accurately matching various information and languages of the terminology data). If an existing translation result is retrieved, the terminology data is associated with the corresponding data in the translation text table through its primary key. If no existing translation result is retrieved, the terminology data pair is deduplicated and used as the terminology to be translated.
[0072] 3. The internationalization development management platform supports translation using AI (i.e., the aforementioned translation model): For example, the internationalization development management platform can use a scheduled task based on Xxljob (a distributed task scheduling system) to execute the translation of terms from a shared terminology library using AI every N minutes (the time interval can be modified in Xxljob).
[0073] For example, the internationalization development management platform can use a scheduled task based on Xxljob (a distributed task scheduling system) to write the AI translation results and their corresponding terms into the translation text table and terminology table every M hours to pre-store them in a shared terminology library (the shared terminology library will be synchronized twice: the first time for the AI translation results, and the second time for the translation results after the proofreading task is completed). The AI translation results pre-stored in the shared terminology library can meet the basic and / or emergency needs of the business system.
[0074] The AI engine, AIGC, is a generative AI engine capable of multilingual translation. For example, AI can be a deeply optimized variant of the AIGC engine and the Transformer-XL architecture, capable of multi-stage training with hundreds of billions of tokens on a dedicated computing cluster (80×NVIDIA A100 GPUs): Stage 1 (General Corpus Pre-training): Constructing the AI's basic semantic representation using a parallel general corpus set covering multiple languages (e.g., 52 core languages); Stage 2 (Domain-Adaptive Fine-tuning Training): Injecting translation result datasets of translated terms from various vertical domains (e.g., machinery / finance / logistics), historical translation memory from internationalization development and management platforms, and / or imported terminology resources from various business systems to strengthen professional context modeling; Stage 3 (Human Feedback Reinforcement Learning (RLHF)): Injecting translation results from completed proofreading tasks to drive AI-generated results to align with professional translator standards.
[0075] For example, the dynamic optimization process when AI uses a terminology database for domain-adaptive fine-tuning training: (1) Real-time terminology injection: When a translation request is triggered, the system automatically associates related entries (e.g., related terms) that have contextual relevance in the terminology database and constrains the model output; (2) Conflict resolution mechanism: When there is a difference between the translation result generated by AI and the terminology database, a confidence-weighted voting is initiated (a weighted voting is conducted based on the confidence of the translation model and the priority of the terminology database). (3) Closed-loop knowledge accumulation: The translation results obtained from completing the proofreading task are automatically written back to the terminology database, triggering AI incremental training (e.g., weekly full fine-tuning + real-time LoRA (Low-Rank Adaptation, an efficient model fine-tuning method)). Among them, the AI translation results obtained by AI translation (i.e., the first translation result) can be deduplicated and put into the term translation library (i.e., the aforementioned proofreading library).
[0076] 4. The administrator of the international development management platform regularly releases proofreading tasks based on the term translation database, so that proofreaders can accept the orders, complete the translation, and obtain the proofreading translation results (i.e., the aforementioned second translation result).
[0077] For example, the administrator of the internationalization development management platform can publish translation tasks every Monday. These tasks select AI-translated terms from a terminology translation library to form proofreading tasks, which are then published via a crowdsourcing mechanism. The number of terms in a single proofreading task can be freely set by the administrator when editing the task; for example, 200 terms could be selected from the terminology translation library for a single task. Under this crowdsourcing mechanism, proofreaders with available time can bid for these tasks on the corresponding client of the internationalization development management platform (i.e., the aforementioned proofreading client). After bidding, they can export the terms for the acquired task. After completing the proofreading, the proofreaders import the translation results back into the terminology translation library, awaiting synchronization back to the shared terminology library. Furthermore, under this crowdsourcing mechanism, if no one bids for a task within a preset time (e.g., one day), the internationalization development management platform randomly assigns proofreading tasks to the client of proofreaders specializing in a particular language.
[0078] For example, the internationalization development management platform can execute incentive controls corresponding to the review end based on preset rules (or settlement rules, which can be configured on the internationalization development management platform) and task completion data. For example, when each review task is completed within a specified time, the fee is calculated at 0.5 yuan per entry, and when it is completed after the time limit, the fee is calculated at 0.3 yuan per entry. The settlement method is determined according to the personnel attributes of the reviewers, and settlement control is carried out according to the determined calculation method. For example, if the personnel attribute of the reviewers is internal personnel, the settlement method is monthly settlement; if the personnel attribute of the reviewers is external personnel, the settlement method is settlement on a per-item basis.
[0079] For example, the high-quality proofreading and translation results obtained from the proofreading task can be added to the terminology database to update the terminology database. Subsequently, the newly added proofreading and translation results in the terminology database can be used to incrementally train the AI to iteratively improve the AI's translation performance.
[0080] 5. The internationalization development management platform provides two methods for business systems to use the translation results corresponding to each term in the shared terminology library: (1) Send an HTTP request (i.e., the aforementioned translation retrieval request) using the business system ID + internationalization key + language to obtain one or more specified translation results. The first request reads from the shared terminology library and caches it in the Redis database. Subsequent requests read the cached data from the Redis database, greatly reducing response time.
[0081] The Redis database sets the cache validity period (i.e., the preset time limit) to 12 hours. After the cache expires, it needs to be retrieved again from the shared term library.
[0082] For example, after a business system obtains the translation results from the Redis database, if the business system modifies the translation results to obtain updated translation results, it can use the API interface configured in the internationalization development management platform to reset the Redis database cache. This allows the updated translation results obtained by the business modules to be reloaded individually or in batches into the Redis database cache according to a preset filtering range, so that each business system can immediately call the latest translation results as needed.
[0083] (2) Export all or part of the data from the shared terminology library through the business system, and store the exported data locally in the business system to achieve local backup of the data. In the event of a failure in the internationalization development and management platform, the business system can ensure that it can obtain the translation results normally through the locally backed-up data.
[0084] For example, each business system can set data acquisition rules, such as prioritizing the acquisition of translation results from the shared term library via on-demand HTTP requests, and if the request fails, acquiring the required translation results from the term resources backed up locally by the business system.
[0085] The technical solution in this example constructs a multi-business system translation service engine based on intelligent scheduling and a shared terminology library. Its purpose is to: (1) design a cross-business system translation resource sharing engine, actively aggregate translation resources (terminology library, translation memory library, and manual proofreading results) of each tenant / business system through the international development management platform, and construct a shared terminology library. The global translation resource reuse of multiple tenants / business systems is realized through intelligent matching algorithms (exact matching / fuzzy matching), eliminating duplicate translations; (2) develop dynamic language configuration to support each tenant / business system to independently configure the required language list, and the international development management platform dynamically expands or deactivates languages as needed; (3) realize the closed-loop optimization of AIGC engine translation and manual proofreading, and provide a low-cost emergency translation channel through AIGC engine translation. Among them, AIGC engine also establishes a human-machine collaboration feedback loop: the manual proofreading translation results are used to reverse train the AI model and update the resource pool to continuously improve the translation quality; (4) build a crowdsourced proofreading architecture; (5) adopt Redis caching to build a high-performance resource caching and invalidation system.
[0086] Specifically: This example's technical solution, through an internationalization development and management platform, addresses the technical problems of fragmented management, redundant construction, low reuse rates, high costs, and wasted manual proofreading resources (including terminology databases, translation memories, and human proofreading results) in a multi-business system, multi-project parallel development environment. It solves the technical shortcomings of existing internationalization solutions, such as isolated translation resources, fragmented management, and high reuse barriers, significantly improving translation efficiency, quality, and economic benefits. Furthermore, through the dynamic language management engine, AIGC deep integration optimization mechanism, and crowdsourced proofreading and deduplication mechanism within the internationalization development and management platform, it achieves efficient reuse of translation resources, significant reduction in translation costs, continuous improvement in translation quality, and enhanced translation compatibility between systems. This example's internationalization development and management platform can significantly improve the translation efficiency and quality consistency of global business operations. Specific details are as follows: 1. Break down resource silos and achieve efficient reuse of translation resources across business systems. Aggregate the terminology resources of existing internationalization solutions scattered across various business systems into a shared terminology library, and achieve efficient reuse of translation resources across business systems through intelligent matching algorithms. Compared to existing internationalization solutions, this directly reduces repetitive translation workload by more than 70%, significantly reduces human and computing power consumption, and increases the coverage of translations for less common languages by 300%. 2. Supports configuring languages independently for business systems (e.g., tenant's business systems) based on their needs (e.g., a list of required languages), flexibly adjusting (e.g., disabling) and adding new languages to achieve dynamic language configuration; 3. Utilizing the AIGC engine for optimization and iteration, translation quality is improved. Translation accuracy in specialized fields is improved by 12.7% (compared to the general GPT-4); terminology consistency reaches 99.4% (ISO / TS 11669:2017 standard test); millisecond-level response is supported for 52 languages (average latency 68ms@P99); furthermore, an AIGC-human closed-loop optimization strategy is established to train the AI and update the resource pool through human proofreading and translation results, continuously improving translation quality. The AIGC engine also enables the provision of a low-cost emergency translation channel. 4. By using crowdsourcing and deduplication mechanisms, we ensure that identical content only requires one manual review. The review and translation results are automatically synchronized to all related business systems through a shared terminology library. This reduces the cost of manual translation by 80% and increases the response speed of emergency translation by 10 times. It enables dynamic resource scheduling and on-demand services, optimizing costs. In addition, the crowdsourcing and deduplication mechanisms also enable the construction of a crowdsourcing network of part-time translators, supporting the distribution of review tasks and incentive settlement. 5. Providing open interfaces for various translation service types (such as AI initial translation / rapid proofreading through crowdsourcing / professional human proofreading, etc.) improves translation efficiency for new languages by 90% and shortens delivery cycles for customized translation scenarios. Providing open interfaces for various translation service types enables a highly scalable architecture to support agile business iteration. 6. By using Redis to cache frequently accessed translation resources and binding expiration tags (such as terminology validity period and model version) to these resources, automatic expiration and updates are implemented (setting lifespan for translation resources and reloading expired resources from the Redis cache). This ensures that translation request response time is less than 50ms (for thousands of TPS of concurrent requests) and resource update latency is less than 1 second. Therefore, the aforementioned high-performance caching ensures support for large-scale concurrent access. Thus, the technical solution in this example uses a Redis caching layer to store frequently accessed resources (terminology database, frequently accessed translation memories), responds to requests in milliseconds, and constructs a high-performance resource caching and expiration system.
[0087] Based on the same inventive concept as the foregoing embodiments, this application provides a computing device, such as... Figure 4 As shown, the device includes: a processor 310 and a memory 311 storing a computer program; wherein, Figure 4 The processor 310 shown in the diagram does not indicate that there is only one processor 310, but only indicates the positional relationship of the processor 310 relative to other devices. In practical applications, there can be one or more processors 310; similarly, Figure 4 The memory 311 shown in the diagram has the same meaning, that is, it is only used to indicate the positional relationship of memory 311 relative to other devices. In practical applications, there can be one or more memories 311. When the processor 310 runs the computer program, the software internationalization processing method applied to the above-mentioned device is implemented.
[0088] The device may also include at least one network interface 312. The various components of the device are coupled together via a bus system 313. It is understood that the bus system 313 is used to implement communication between these components. In addition to a data bus, the bus system 313 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 4 The general designated all buses as Bus System 313.
[0089] The memory 311 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memory 311 described in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.
[0090] The memory 311 in this embodiment is used to store various types of data to support the operation of the device. Examples of this data include: any computer programs used to operate on the device, such as operating systems and applications; contact data; phonebook data; messages; pictures; videos, etc. The operating system includes various system programs, such as the framework layer, core library layer, driver layer, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications, such as media players, browsers, etc., used to implement various application services. Here, the program implementing the method of this embodiment can be included in the application.
[0091] Based on the same inventive concept as the foregoing embodiments, this embodiment also provides a computer-readable storage medium storing a computer program. The computer-readable storage medium can be a magnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc. When the computer program stored in the computer-readable storage medium is executed by a processor, it implements the above-mentioned software internationalization processing method. For the specific steps implemented when the computer program is executed by the processor, please refer to [link to relevant documentation]. Figure 1 The description of the illustrated embodiments will not be repeated here.
[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0093] In this document, the terms “including,” “comprising,” or any other variations thereof are intended to cover non-exclusive inclusion, which includes not only the elements listed but also other elements not expressly listed.
[0094] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method of software internationalization processing, characterized by, The method comprises the steps of: S11: obtaining a to-be-translated term in a shared term library, wherein the translation result in the shared term library is used by multiple business modules; S12: calling a translation model to translate the to-be-translated term to obtain a corresponding first translation result, and adding the first translation result to the shared term library; S13: entering a review process according to the to-be-translated term or the first translation result to obtain a corresponding second translation result, and synchronizing the second translation result to the shared term library.
2. The software internationalization processing method of claim 1, wherein, The review process comprises: adding the to-be-translated term or the first translation result to a review library; creating one or more review tasks based on the review library, wherein the review tasks include a preset number of to-be-translated terms or first translation results; task publishing the review tasks to send corresponding task notifications to one or more reviewers; receiving at least one second translation result fed back by the reviewer after accepting and completing the review task, wherein the reviewer completes the review task in a manner comprising manual review and / or model review; and / or recording task completion data of the reviewer, and / or performing incentive control corresponding to the reviewer according to a preset rule and the task completion data.
3. The software internationalization processing method of claim 2, wherein, The adding of the to-be-translated term or the first translation result to the review library comprises: repeated detection of the review library according to the to-be-translated term or the first translation result; when the repeated detection result represents no repetition, adding the to-be-translated term or the first translation result to the review library; or when the repeated detection result represents repetition, performing a de-duplication process on the review library.
4. The software internationalization processing method of claim 2, wherein, After step S13, comprising: triggering incremental training of the translation model based on the second translation result to fine-tune the translation model.
5. The software internationalization processing method of claim 1, wherein, In step S12, comprising: obtaining a language requirement; obtaining associated terms in the shared term library that are contextually associated with the to-be-translated term; calling the translation model for translation according to the to-be-translated term, the associated term, and / or the language requirement to obtain the first translation result corresponding to the to-be-translated term, wherein the language requirement corresponds to one or more languages, and the translation model is a generative model using AIGC technology.
6. The software internationalization processing method according to any one of claims 1 to 5, characterized in that, Further comprising: receiving term resources imported by the business module, and adding at least part of the data in the term resources to the shared term library, wherein the at least part of the data includes to-be-translated terms and / or translation results of translated terms.
7. The software internationalization processing method according to any one of claims 1 to 5, characterized by, Further comprising: in response to a translation acquisition request of the business module, when the translation acquisition request is a first request within a preset time limit, obtaining a corresponding translation result from the shared term library and sending it to the business module, and storing the translation result corresponding to the translation acquisition request in a Redis cache according to the preset time limit, or when the translation acquisition request is a non-first request within the preset time limit, obtaining a corresponding translation result from the Redis cache and sending it to the business module; and / or, According to preset backup rules, some or all of the translation results in the shared thesaurus are backed up to the business module.
8. The software internationalization processing method according to any one of claims 1 to 5, characterized by, The method for building the translation model includes: The model framework is trained using general corpora corresponding to multiple languages and / or terminology databases corresponding to multiple preset domains to obtain the initial translation model; and / or, The translation model is trained using the translation results of translated terms from the term resources imported by the business module, in order to update the translation model.
9. A computing device, comprising: include: A processor and a memory storing a computer program, wherein, when the processor runs the computer program, the steps of the software internationalization processing method according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the software internationalization processing method according to any one of claims 1 to 8.