Pushing The Translation Quality Envelope Further

进一步提高翻译质量

2020-06-09 22:50 Wordbee Translator

本文共776个字,阅读需8分钟

阅读模式 切换至中文

Translation quality is an evergreen topic as well as a pain point within the localization industry. In the Wordbee blog, we’ve already tackled this subject with two articles about the right terminology to use when talking about quality and a holistic approach to quality evaluation. In this article, we want to talk about why it makes sense to invest in translation quality measurement and how companies can benefit from the collaboration between Wordbee and ContentQuo. ContentQuo is an Estonia-based translation technology startup serving a few Global Top-10 LSPs and several corporate localization departments. Based on the request of a global enterprise client, ContentQuo developed an integration to connect their translation quality management technology with the Wordbee translation management system in 2019. To understand how ContentQuo can integrate in Wordbee and what its benefits are, we met virtually with its founder and CEO, Kirill Soloviev. Here is what he had to say. Investing in translation quality Kirill Soloviev founded ContentQuo in 2015 drawing on his 18+ years in the language industry as a translation buyer. ContentQuo aims at improving the collaboration between buyers and vendors as the key to increase translation quality. Nevertheless, collaboration requires quality measurement and investment in the associated technologies. As an investor, you don’t want to put all your eggs into one basket; you want to spread your budget into different buckets to attain the maximum return with the minimum risk. “The same goes for translation quality”, explains Soloviev. “Look at your budget and consider carefully all the languages and content streams: where do you want to deploy your investment to reduce the risk of losing quality?” A translation assembly line A translation organization is very similar to an assembly line. Soloviev: “You take your content, usually in one language, and enter it in the translation factory. Different people in the organization work on the content, revising it, translating it, adapting it while the translation is still ongoing. At the end of the line, the content is localized in as many languages as you’ve planned.” Just like in most assembly lines, the end product is usually compliant to set criteria. But, sometimes, something can go wrong along the assembly line. The outcome can then be flawed, i.e. contain some minor or major errors, or not be fit for the target group. With the help of a quality measurement tool, a quality manager can look at every single step of the localization process: inspect the deliverables and detect if something went wrong; fix the problem and get everything right the next time. Or, even better, detect a potential risk, remove it and get everything right from the start. Methodology agnosticism There is no one-standard-fits-all approach to quality, says Soloviev: “We’re not here to change the way people measure translation quality (unless they ask us to); we want to help localization professionals measure it in the way they want - efficiently and effectively.” Each content stream requires a different evaluation method. The quality evaluation of perishable content might require a time-saving holistic approach (e.g. based on adequacy and fluency), while the evaluation of publishable content is supposed to be analytical and in some cases it might require sub-segment annotation. Translation quality for humans and machines The integration of ContentQuo in Wordbee offers three main opportunities to its users. Vendor managementWhether your company has already the core translation workflow in place and wants to adopt a more sophisticated approach or you’re just starting out with quality evaluation, “this integration will allow you to measure vendor performance and give feedback to vendors to allow you to have the best translation possible (at the given conditions) with the best collaboration.” As a localization manager, by using targeted audits and routine checks, you’ll be able to answer your supervisor’s basic and recurrent question: Are we getting what we are paying for? Use of machine translation. Neural machine translation (NMT) is getting more and more traction. For customers that use NMT (stock or custom engines), “the integration of ContentQuo in Wordbee contributes to monitoring the quality of an NMT engine, to see whether it’s improving, stagnating or degrading. And when it comes to post-editing, you can analyze the post-editing effort and find ways to improve your methodology and NMT system, when customizable.” Continuous improvement. By analyzing the available historical data, you can set your own benchmarks and see whether your translation effort is improving, or, if necessary, monitor translation quality on a daily basis.  For more on the topic of translation quality management, you can listen to Wordbee’s International Buzz podcast interview with Kirill Soloviev.
翻译质量是一个永恒的话题,也是本地化行业的一个痛点。在Wordbee博客中,我们已经通过两篇文章来讨论这个问题,这两篇文章讨论了在讨论质量时应该使用的正确术语以及质量评估的整体方法。在这篇文章中,我们想谈谈为什么投资于翻译质量评估是有意义的,以及公司如何从Wordbee和ContentQuo的合作中获益。 ContentQuo是一家位于爱沙尼亚的翻译技术初创公司,为一些全球排名前10的语言服务提供商和多个企业本地化部门提供服务。根据全球企业客户的要求,ContentQuo于2019年开发了一个将其翻译质量管理技术与Wordbee翻译管理系统相连接的集成。为了了解ContentQuo如何整合到Wordbee中,以及它的好处是什么,我们与它的创始人兼首席执行官Kirill Soloviev进行了虚拟会面。下面是谈话内容。 翻译质量的投资 Kirill Soloviev于2015年创立了ContentQuo,这得益于他作为翻译买家在语言行业工作了18年。ContentQuo的目标是提高翻译质量,加强买家和卖家之间的合作。 不过,合作需要质量评估和对相关技术的投资。作为一个投资者,你不想把所有的鸡蛋放在一个篮子里;你想把你的预算分散到不同的桶里,以最小的风险获得最大的回报。索洛维耶夫解释说:“翻译质量也是如此。”。“看看你的预算,仔细考虑所有的语言和内容流:你想把你的投资部署到哪里来降低质量下降的风险?” 翻译流水线 翻译组织与流水线非常相似。索洛维耶夫:“你把你的内容,通常是一种语言,输入翻译工厂。在翻译过程中,组织中的不同人员会对内容进行修改、翻译、改编。最后,内容将按照您的计划以尽可能多的语言进行本地化。” 正如大多数流水线一样,最终产品通常符合设定的标准。但是,有时候,流水线上会出问题。结果可能会有缺陷,即包含一些小错误或大错误,或者不适合目标群体。 在质量度量工具的帮助下,质量经理可以查看本地化过程中的每一个步骤:检查可交付成果并检测出是否出了问题;解决问题并在下一次将一切都做好。或者,更好的是,发现一个潜在的风险,将其消除,从一开始就把一切都做好。 方法论不可知论 索洛维耶夫说:“我们并不是要改变人们衡量翻译质量的方法(除非他们要求我们这样做);我们希望帮助本地化专业人士以他们想要的方式来衡量翻译质量——高效且有效。” 每个内容流需要不同的评估方法。易腐内容的质量评估可能需要一种节省时间的整体方法(例如,基于充分性和流畅性),而可发布内容的评估应该是分析性的,在某些情况下可能需要子部分注释。 人机翻译质量 Wordbee中ContentQuo的集成为其用户提供了三个主要的机会。 供应商管理。无论您的公司已经有了核心的翻译工作流程,并希望采用更复杂的方法,还是您刚开始进行质量评估,“此集成将允许您衡量供应商绩效,并向供应商提供反馈,以使您能够在最佳合作下(在给定条件下)获得最佳翻译。”作为本地化经理,通过有针对性的审核和例行检查,你将能够回答你的上司的一个基本的和反复出现的问题:我们得到的是我们支付的吗? 使用机器翻译。神经机器翻译(NMT)正受到越来越多的关注。对于使用NMT(库存或定制引擎)的客户,“Wordbee中ContentQuo的集成有助于监控NMT引擎的质量,看它是在改进、停滞还是退化。当谈到后期编辑时,您可以分析后期编辑工作,并在可定制的情况下找到改进方法和NMT系统的方法。” 持续改进。您可以通过分析可用的历史数据,设置自己的基准,查看您的翻译工作是否在完善,或者,如果需要,您可以每天监控翻译质量。 了解更多关于翻译质量管理的话题,您可以收听基里尔·索洛维耶夫在WordBee的国际Buzz播客的采访。

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