基于本地化开发流程优化的机器辅助翻译系统研究与设计

【作者】 李力；

【导师】 高海锋； 李品翔；

【作者基本信息】 复旦大学 ， 项目管理（专业学位）， 2013， 硕士

【摘要】 随着世界经济,技术的快速发展,软件行业处在高速的发展变化时期。经济的全球化促进了软件产业的国际化,软件的国际化生产和全球服务成为国际软件公司的基本发展策略；于此同时软件开发过程中的工具和流程也已经经历了多次的变革和进步,为了提升市场份额,支持全球化软件开发和升级速度也大大的提高。然而支持全球化的软件产品的开发渐渐成了软件开发的一个瓶颈。一个软件产品往往需要支持几种到几十种本地化语言。对于不同语言的翻译,编码方式的转换,度量衡单位的变化,以及对多操作系统和不同系统建构的支持成为很多软件开发者的梦魇。传统的本地化软件的开发模式是在源语言版本的软件上进行的二次开发,提升国际化的支持和本地化语言翻译的内容,但是这种模式必然会带来开发周期的延长和成本的上升。为了达到缩短本地化版本开发周期的目的,需要采用更加敏捷的流程和新的自动化工具提高开发效率和产品质量。流程的优化需要工具的配合,工具的集成帮助流程更好的执行。本文研究的内容侧重于机器辅助翻译软件系统的架构和算法的实现以及由此带来的全球化软件项目中开发流程的优化。针对本地化开发过程中的问题,本文主要解决了三个个核心的问题：首先与集成编译平台整合：自动化的抓取源语言内容,加以翻译,并自动生成编译系统所需要的本地化资源或代码,从而定期的获得包含最新翻译内容的本地化版本,从而更早发现本地化的软件缺陷；其次经过利用实例翻译和翻译记忆对源语言资源进行翻译之后,机器翻译的对照表可以与人工翻译的系统对接,人工校验和更正可以直接反映在下一迭代生成的本地化版本中；最后系统的升级与整合在管理上提供流程上的进一步优化的空间,优化关键路径上的任务,减少任务之间的依赖关系,从而提升生成效率和产品质量。更多还原

【Abstract】 Software has changed dramatically over the past decades. The globalization of the economy promotes the globalization software industry. Globalization becomes the basic strategy in this trend. The demand for software products in more and more languages exploded. Shorter delivery cycle will be able to help to gain more market share. However, globalization is the bottleneck in the software development. A globalized product needs to support several dozens of language. Localization team has to face more and more challenges like encoding conversion, measurement unit and supportability for multiple operating systems and system framework. The traditional localized software development model mostly likes developing a new version with internationalization and localization capabilities, but this model is bound to bring the extension of the development cycle and cost rise. We need a more agile process and new automation tools to improve efficiency and quality.This study focused on architecture design of Computer Assisted Translation system, translation algorithm and optimization of globalization software development process. The dissertation is mainly to solve three core problems: Firstly, Integrated with Compiler Integration platform, the content in source language is fetched out and translated automatically, the target content is injected into Source Control System and generated localized build regularly. Developer and tester can earlier found localization defects than ever; Secondly, based on the result of the Instance Translation and Translation Memory Translation, the translation engineer can import the result into standard translation tools and write the updates back to CI system to generate new localized version; Thirdly The system provides further optimization space in process. Optimization tasks on the critical path and reducing dependencies between tasks, it helps to enhance efficiency and product quality. It makes more agile to be possible.更多还原