Four Questions to Guide Your Localization Technology Strategy

本地化技术四问

2020-03-03 15:30 RWS Moravia Insights

本文共1109个字,阅读需12分钟

阅读模式 切换至中文

“What technologies do I need to power my localization program?” It’s a common question among global businesses just getting started with localization, but it’s a little too broad. There are so many tools that can enhance your program, but you first need to know what needs enhancing. A better place to start would be with more specific questions around the problem areas you’d like localization technology to solve. Those are the kinds of questions we’ll explore here. You’ll learn all about the basic “starter pack” of localization technologies (as we like to think of it) before we touch on the latest tech developments. First, for a little context, you should know about the mother of all localization technologies: the translation management system (TMS). This is the platform that stitches together all the tools companies use to manage their translations. It will be the core of your technology stack, and requires serious planning before you buy. (Here’s an ebook that will help guide how to choose one.) All other localization technologies you’ll hear about—some of them components of TMS—tend to fall into one of three categories: Computer-assisted translation (CAT) tools Machine translation (MT) tools Linguistic quality assurance (LQA) tools Here are examples of common problems you might be trying to solve. Even more specific questions might be: How do you help your translators reuse work that has already been done? Or how do you make sure the translator has the right context and the information they need? For this, you’ll need CAT tools, which are a central component of any TMS. Specifically, they have three functions that can help you in your efforts: Translation Memory (TM). A TM is a database of completed translations the translator can pull from so that when they are translating something identical or very similar to a past translation, they don’t have to reinvent the wheel. Terminology database. Not only might you have repeat content, but you likely use industry terminology. Most CAT tools include a terminology database that provides translations plus information around how and when to use these terms. Visual context, also known as in-context review, is available for some, but not all, file types and/or CAT deployments. When it is available, translators can see the format in which the content will be published as they work. Using one or more of these tools helps control content consistency across formats and languages. Automation can help translators work faster, but must be carefully applied if you’re going to produce content fast and to a high standard. In this case, you want to think about the components of a TMS that can automate your workflows. One option, again, is to use translation memory to pull past translations. Say you’re updating a product manual you’ve translated before: your TM should automatically recycle a large portion of that content, leaving time-consuming manual work behind. Another option, which you might apply to the remaining untranslated content after the TM has done its work, is machine translation (MT). MT engines can translate text without human intervention—think tools like Google Translate. At its current stage of advancement, MT can handle basic, informational content like knowledge-base articles, user manuals and customer chat bots. Typically, you’ll want to take a hybrid approach and combine MT with human post-editing, unless the content is of limited and temporary use. (For persuasive content that carries emotional weight, you’ll probably want to forego MT for now.) As a bonus, TM and MT don’t just make humans more efficient. They also provide inspiration. Past translations can clue translators in to better ways to translate things in terms of verbiage or client preferences. Which is better for translators to use—TM or MT? Check out this post to dive deeper. This is where the third category of localization technology comes in: linguistic quality assurance (LQA) tools. These tools flag basic errors like spelling or punctuation mistakes, inconsistent capitalization, extra spaces and terminology mismatches for the human reviewer to clean up. They speed up work and eliminate careless human errors, but humans are still needed to make the final edits, since the technology isn’t advanced enough to fully rely on yet and still generates false positives. You can also set up customized rules-based QA tools to check for certain types of phrasing that might be problematic to your target market, like geopolitical statements with mentions of contested territories. QA tools can help catch cultural mistakes that have big implications for your brand. LQA tools can also integrate with your TMS so linguists can access them from within the translation environment. A company expanding into Europe might not want to invest a lot of money in Greece in the wake of its economic crisis. But that’s not to say Greece is completely unviable. To gauge consumer interest in your product in new markets like these, you could use a combination of MT and crowdsourcing. Crowdsourcing platforms (sometimes deployed as a superset of CAT tools) tend to offer lower-cost translations, so there’s less investment at stake. You start with MT, then, through crowdsourcing platforms—accessible by “crowds” of translators who all work on small pieces of your project simultaneously—you can leverage multiple native-language speakers to suggest improvements to your machine-generated translations rather than using more expensive (though more reliable) professional post-editors. The danger is, you release sub-par and potentially inconsistent translations, but it’s not a bad move in markets where any risk of a tarnished brand footprint will be smaller. (Do not try this in a major market.) You’ve probably heard of artificial intelligence, and yes, AI is disrupting localization as much as any other industry. The above technologies will get you started, but as your localization program matures, you might find yourself asking questions that can only be answered with AI-powered solutions. For example: How can we route jobs to best-fit translators? (Crowdsourcing platforms are advancing to automate this process.) How can we track and manage the costs of many small transactions? (AI can help build powerful dashboards.) How can we perform sentiment analysis on multilingual content? (Google has developed tools like the Natural Language API to classify and analyze speech.) How can we machine-translate persuasive content? (Nothing on the horizon here. We’ll just have to wait and see if MT is eventually up to it.) In the meantime, have a think about the business problems you need localization technology to solve, and use those to guide your buying choices.
“我需要什么技术来支持我的本地化项目?” 这是全球性企业在刚开始进行本地化时会提出的一个常见问题,但这个问题有点太宽泛了。事实上,有很多工具可以帮到您,但是您首先得知道您的项目需要提升的是什么。因此,您最好考虑一下,您希望本地化技术可以解决哪方面的问题,然后再据此提出更具体的问题。 这就是我们要在这篇文章中探讨的内容。在了解最新技术的发展之前,您首先需要了解本地化技术的基本“入门包”。 首先,您应该了解所有本地化技术的基础:翻译管理系统(TMS)。这是一个将所有翻译管理工具整合在一起的平台。它是所有技术工具的核心,因此,在购买本地化系统之前,您需要认真的规划。 所有其他本地化技术(其中一些是TMS的组件)都可以归纳为以下三种类型: •计算机辅助翻译(CAT)工具 •机器翻译(MT)工具 •语言质量保证(LQA)工具 下面是一些关于本地化的常见问题,有些或许正是还在困扰您的问题。 问题1:如何使翻译更一致? 确切来说,应该是:如何让您的译者重新使用过去的工作成果?或者如何确保译者了解正确的语境和他们所需信息? 为此,您需要计算机辅助翻译(CAT)工具,它是任何一种TMS的核心组件。具体来说,它有以下三个功能: 翻译存储(TM)。TM是一个可供译者提取的已完成翻译数据库,当译者翻译到重复或非常相似的内容时,他们就不必重复工作了。 术语库。译者在翻译中不仅会遇到重复的内容,也很可能会遇到某些行业的专业术语。大多数CAT工具都囊括术语库,它可以为译者提供术语翻译以及指导译者使用这些术语的方式和时间。 可视上下文(也称为上下文内审查)。这个功能可应用于某些特定文件和(或)CAT的部署。通过该功能,译者可以查看未来译文发布时的格式。 使用其中一个或多个工具有助于译者统一格式和保证翻译的一致性。 问题2:如何减少创建多语言内容所需的时间? 自动化可以帮助译者更快地工作。但如果您想快速、高质量地生成内容,则必须谨慎对待自动化。 在这种情况下,您需要考虑可以使您的工作流自动化的TMS组件。 其中一种方案是使用翻译存储(TM)来提取过去的翻译。假设您正在更新以前翻译过的产品手册,您的TM便会自动更新其中的大部分内容,这样可以大大减少耗时的人工工作量。 另一种方案是机器翻译(MT)。您可以用其处理机器翻译后剩余的未翻译内容。机器翻译软件可以在无译者介入的情况下自动翻译文本,如谷歌翻译。就目前的发展阶段来说,机器翻译可以处理一些基本的信息,如科普文章、用户手册和智能交互聊天机器人。一般来说,除非译本有限或只是临时使用,人们会采用一种折中的方法,即把机翻和译者后期编辑结合起来(带有情感色彩的说服性文章一般不会选用机器翻译)。 TM和MT不仅可以提高工作效率,也可以为译者提供灵感。过去的翻译可以为译者提供更好的翻译方法,比如措辞和客户偏好。 问题3:如何检验译者工作的质量? 这就要涉及到本地化技术的第三类应用:语言质量保证(LQA)工具。 这些工具可以标记基本错误,如拼写或标点错误、大小写不一致、额外的空格和术语不匹配等,以便审校人员进行修正。它们有助于加快工作速度,消除粗心大意的人为错误,但是,最终译文仍然需要人工进行最后的编辑,因为目前该技术还不够先进,不是完全可靠,而且可能会有误判。 您还可以自定义一些质量保证(QA)工具来检查某些特定措辞。这些措辞一旦出问题,可能会给您带来麻烦,比如地缘政治声明中提到的有争议的领土。QA工具可以检查出对您的品牌有重大影响的文化性错误。 LQA工具还可以与TMS集成,这样语言学家就可以在翻译语境内使用它们。 问题4:如何快速且经济地测试市场生存能力? 在希腊经济危机之后,一家已经扩张到欧洲的公司便可能不想在希腊投入大量资金。但这并不是说开拓希腊市场是完全不可行的。要衡量新市场消费者对您的产品的兴趣,您可以使用机器翻译与众包结合的方法。 众包平台(有时作为CAT工具的超集)往往能提供更低成本的翻译,因此投资风险更小。您一开始需要使用机器翻译,然后通过众包平台,即一个可以让多位译者同时编辑的平台,借助多个原生语言使用者的建议改进机器生成的翻译,而不需要雇佣更昂贵的专业译后编辑(虽然他们可能更可靠)。 这其中的风险是,您发布的译文质量可能不合格,而且语言内容可能不一致,但这在那些品牌形象受损风险较小的市场上来说并不完全是一个糟糕的尝试(但是不要在主流市场尝试这种做法)。 二级技术 您可能听说过人工智能,是的,人工智能对本地化的影响不亚于任何其他行业。上述技术将帮助您入门,但是随着您的本地化项目要求越来越高,有些问题只能通过AI来解决。例如: ·如何将工作安排到最合适的译员手中?(众包平台正在推进这一过程的自动化) ·如何跟踪和管理小额交易的成本?(人工智能可以帮助构建强大的交互面板) ·如何对多语言内容进行情感分析?(谷歌开发了像“自然语言API”这样用来分类和分析语音的工具) ·如何用机器来翻译说服性内容?(在这一方面目前什么进展也没有,我们只能等着看机器翻译最终是否能胜任) 同时,您还要考虑需要运用本地化技术来解决的业务问题,请根据以上内容来选择您想要购买的本地化系统。 译后编译:陆遥 (中山大学)

以上中文文本为机器翻译,存在不同程度偏差和错误,请理解并参考英文原文阅读。

阅读原文