The Artistry of Post-Editing


2024-04-11 02:00 CSOFT


阅读模式 切换至中文

In today’s very globalized world, and amid unprecedented technological progress, the need for seamless communication across languages has never been greater. Machine Translation technology has, therefore, become a mainstream practice, and it has undergone a remarkable transformation, fueled by advancements presenting users with important features, such as consistency, speed, cost reduction, and productivity. But the old question remains – is it utterly successful? By now, everyone working in the translation industry should know what MTPE (Machine Translation Post-Editing) is. Simply, it’s an approach that combines machines’ computational power with the nuanced understanding of human linguists. The machine has advanced to a point where it can deliver cost-effective human-like first-draft translations. But yes, you read it right. It’s a first draft, meaning it’s not quite there yet. The finesse and cultural nuances of language often escape this system that struggles to capture context accurately, leading to a loss of meaning and awkwardness in the target language. And this is where humans remain relevant. They’re there to carefully dot the I’s and cross the T’s, ensure the language is as smooth and readable as possible, and fix what remains beyond the machine’s grasp. Human understanding and enhancement are indispensable to the success of this technology, and post-editing makes the workflow complete and valuable. Here’s an example in which the machine could not flag the error in the source text and figure out that the sentence was incomplete. It translated the English verb can (to be) into Portuguese as a noun — “Latas,” which line the shelves at your local supermarket’s canned goods section. It took the understanding of a human linguist to have the error flagged in the source text so the translation could be edited accordingly. Just to give the machine the benefit of the doubt, I’d like to think that it was trying to save us some money by pushing us to buy canned food. Two types of post-editing are the most common: LPE (Light Post-Editing), which involves minimal human intervention for less critical tasks where only basic, obvious errors and stylistic improvements are looked for, and FPE (Full Post-Editing), which requires very humanized translation and needs extensive human intervention to correct errors and improve fluency. If the translation needs to meet high standards – with, for example, minimal errors and fluent language — FPE may be more appropriate. Strong cases for LPE would be for scenarios where the translation is of internal or informal communications. This is why linguists and project managers are so valuable to the translation industry. They know which type of post-editing each project would benefit the most from, provided they step up at this stage and determine the most appropriate post-editing type, whether LPE or FPE, to achieve the desired balance between quality, time, and cost. What’s the purpose of the translation? What’s the level of quality needed? Are there any time constraints and/or budget considerations? These applications are often industry-specific, and a business’s work determines which approach best suits its needs. For instance, legal, technical, and financial translations often require FPE as the content is typically complex and prioritizes accuracy to properly communicate specifics to its intended audience. On the other hand, marketing translations require a balance between accuracy and creativity, and it’s necessary here to accommodate cultural nuances in the target language. What’s very important is to be able to define the quality criteria upfront and tailor post-editing solutions to the specific client needs and requirements of any project, and that’s what we do over at CSOFT. As a leading provider of global translations, CSOFT adopts an approach that relies on project assessment and management. We understand clients’ expectations and work closely with them in the initial stages to determine their priorities regarding translation quality, turnaround time, and budget. Therefore, a set of sequential stages is involved for any project, such as assessing its needs and requirements, defining the area of expertise, target audience, and time needed for post-editing, evaluating the Machine Translation output, and assembling the adequate translation team accordingly. CSOFT’s MTPE project management team has developed this strategic workflow to maximize automation while ensuring high-quality outputs. As a linguist, I don’t like to think that machines and language experts are eternal rivals. Machines are just tools — advanced technological “extra hands” that can assist the experts and be more productive. Machine Translation Technology complements our linguists’ skills and expertise, providing them with a powerful tool to streamline the translation process. And while we at CSOFT keep our eyes open to new technological developments and how we can use their power to better serve our clients, we know that human linguists – our linguists – will need to work their magic to nail cultural nuances and deliver flawless texts, at least for the foreseeable future. Without a guiding hand, machine translation has the potential to open a lata of its own for its users – one full of worms.
在当今高度全球化的世界中,在前所未有的技术进步中,对跨语言无缝沟通的需求从未如此强烈。因此,机器翻译技术已经成为一种主流实践,它经历了一个显着的转变,这是由为用户提供重要功能的进步推动的,例如一致性,速度,成本降低和生产力。但老问题仍然存在-它是完全成功的吗? 到目前为止,每个从事翻译行业的人都应该知道MTPE(机器翻译后编辑)是什么。简单地说,这是一种将机器的计算能力与人类语言学家的细微差别相结合的方法。这台机器已经发展到可以提供具有成本效益的类似人类的初稿翻译的地步。但没错你没看错这是第一稿,意味着它还没有完全完成。 语言的微妙之处和文化上的细微差别往往会逃脱这个努力准确捕捉语境的系统,导致目标语言失去意义和尴尬。这就是人类仍然相关的地方。他们在那里仔细地点我的和交叉的T的,确保语言是尽可能流畅和可读的,并修复什么仍然超出了机器的掌握。人类的理解和增强对于这项技术的成功是必不可少的,后期编辑使工作流程完整且有价值。 这里有一个例子,在这个例子中,机器无法标记源文本中的错误,并指出句子不完整。它将英语动词can(to be)翻译成葡萄牙语的名词--“Latas”,它排列在当地超市罐头食品区的货架上。这需要人类语言学家的理解力,才能在源文本中标记出错误,以便对翻译进行相应的编辑。为了给机器带来好处的怀疑,我想它是想通过推动我们购买罐头食品来节省我们的钱。 两种类型的后期编辑是最常见的:LPE(轻度后期编辑),它涉及最少的人为干预,用于不太关键的任务,只寻找基本的,明显的错误和文体改进,以及FPE(完全后期编辑),它需要非常人性化的翻译,需要大量的人为干预来纠正错误并提高流畅性。如果翻译需要达到高标准-例如,最小的错误和流畅的语言- FPE可能更合适。LPE的强有力的案例是翻译是内部或非正式沟通的情况。 这就是为什么语言学家和项目经理对翻译行业如此重要。他们知道每个项目最受益于哪种类型的后期编辑,只要他们在这个阶段加强并确定最合适的后期编辑类型,无论是LPE还是FPE,以实现质量,时间和成本之间的理想平衡。翻译的目的是什么?需要什么样的质量水平?是否有任何时间限制和/或预算考虑? 这些应用程序通常是特定于行业的,企业的工作决定了哪种方法最适合其需求。例如,法律、技术和金融翻译通常需要FPE,因为内容通常很复杂,并且优先考虑准确性,以便向目标受众正确传达具体信息。另一方面,营销翻译需要在准确性和创造性之间取得平衡,在这里有必要适应目标语言的文化细微差别。非常重要的是能够预先定义质量标准,并根据任何项目的特定客户需求和要求定制后期编辑解决方案,这就是我们在CSOFT所做的。 作为全球领先的翻译服务提供商,CSOFT采用依赖于项目评估和管理的方法。我们了解客户的期望,并在初始阶段与他们密切合作,以确定他们在翻译质量、周转时间和预算方面的优先事项。因此,任何项目都涉及一系列连续的阶段,例如评估其需求和要求,定义专业领域,目标受众和后期编辑所需的时间,评估机器翻译输出,并相应地组建适当的翻译团队。 CSOFT的MTPE项目管理团队开发了这一战略工作流程,以最大限度地提高自动化,同时确保高质量的输出。 作为一名语言学家,我不喜欢认为机器和语言专家是永恒的对手。机器只是工具--先进的技术“额外的手”,可以帮助专家,提高生产力。机器翻译技术补充了我们语言学家的技能和专业知识,为他们提供了一个强大的工具来简化翻译过程。虽然我们在CSOFT一直关注新的技术发展,以及我们如何利用它们的力量更好地服务于我们的客户,但我们知道人类语言学家-我们的语言学家-将需要发挥他们的魔力,以确定文化的细微差别,并提供完美的文本,至少在可预见的未来。如果没有一只指导之手,机器翻译有可能为用户打开一个自己的lata-一个充满蠕虫的lata。