What is pre-editing?


2019-11-21 17:54 TranslationMedia


阅读模式 切换至中文

Post-editing machine translation is attracting an increasing interest and many linguists already specialise in providing this service. Corrections made by post-editors are used to re-train MT engines and therefore improve the machine translation output over time. However, post-editing is not the only way in which humans can boost the quality of machine translation. What is pre-editing? Pre-editing happens before a given document is machine translated. Some call it different terms such as “global readiness” or “global readiness editing”. It can involve tasks of varying complexity and ideally is carried out by a skilled editor who is able to analyse the source copy from the perspective of an MT engine. The editor anticipates any potential errors that might occur in the machine translation and amends the source document to make it easier for the engine to translate. This process includes: • reducing sentence length • unifying terminology if it is inconsistent throughout the text • correcting spelling and punctuation errors • simplifying grammatical structures such as subordinate clauses • adding articles and pronouns • removing any ambiguities What’s more, as part of the process, the editor normally makes use of automatic spell-checking and grammar-checking tools. They might also tag any elements that should not be translated. In some cases, pre-editing also means adapting the source copy to closer match the type of content the MT engine has been trained with. Just like post-editing, pre-editing can be done with varying levels of intensity, so it is important to determine what the expected outcome should look like upfront. If carried out correctly, pre-editing can dramatically reduce or, in certain cases, even completely eliminate the need for post-editing. Why consider pre-editing There are a number of reasons why pre-editing a text before it is translated might be a good idea. First and foremost, it is likely to maximise the chances of machine translation output being of good quality and in turn, reduce the time that will be needed for post-editing. The productivity boots can be clearly noted especially when the source document is not well written and it is to be translated into a number of languages. The more target languages there are, the more time will be saved thanks to pre-editing. The above is also directly related to cost savings, something that both language service providers and clients always look for as part of their business strategy. What’s more, pre-editing can alleviate some of the linguists’ frustrations as it saves them from correcting some of the repetitive errors regularly coming up in the machine translation output. Finally, it is also beneficial from an end-user experience. Better localised content means better customer experience. When to pre-edit It might be worth considering pre-editing in large projects with several target languages as time and cost saving will effectively be multiplied by the number of languages. However, it is also advisable to remember that pre-editing is not a magical solution to all machine translation errors. There will be instances where adopting this approach is not absolutely necessary or simply doesn’t offer a good return on investment. When the source text is already of good quality and the machine translation engines have been trained on the domain-specific resources, then pre-editing might not bring a considerable improvement and the time spent on pre-editing will not be offset by shorter post-editing time. Other things to consider The discussion around pre-editing raises a question whether writers of the original source texts should be conscious of the fact that their pieces might be translated or even machine translated at some point in the future and adjust their way of writing accordingly. This MultiLingual magazine article gives some useful tips regarding writing content that is destined for translation. On one hand, it could mean a reduced need for pre-editing, significantly better translation outcomes and time efficiencies. On the other hand, it could limit writers’ creative expression and mean a less enjoyable experience for the target readers of the text in the original language.
编辑后机器翻译吸引了人们越来越多的兴趣,已经有许多人士能专业地提供此项服务。译后编辑人员对译文进行的更改可再次用于训练机器翻译引擎,从而随着时间的推移提高机器译文的水平。但是,后期编辑并不是人类提高机器翻译质量的唯一途径。 何为“预编辑”? 预编辑要在对给定文档进行机器翻译之前进行。“预编辑”有不同的名称,例如“全局准备”或“全局准备编辑”。它可能涉及复杂程度各异的任务,理想情况下,预编辑人员需要从机器翻译引擎的角度分析源文本。预编辑人员预计机器翻译中可能发生的任何潜在错误,并修改源文档以使引擎更容易翻译。 此过程包括以下内容: •缩短句子长度; •如果全文术语不一致,统一术语; •纠正拼写和标点错误; •简化语法结构,例如从句 •添加冠词和代词 •消除歧义 此外,在此过程中,预编辑人员通常会使用自动拼写检查和语法检查工具,还可以标记不需要翻译的语句。在某些情况下,预编辑还要修改源文档,使其更符合机器翻译引擎受训过的内容类型。 就像译后编辑一样,预编辑可以有不同的强度级别,因此确定预期结果的要求很重要。 如果预编辑做得好,可以大大减少、或者在某些情况下甚至完全消除译后编辑的需要。 为什么要进行预编辑? 在翻译文本之前进行预编辑是很好的想法,原因有很多。首先,预编辑很可能会最大程度地提高机器翻译输出质量,从而减少译后编辑所需的时间。尤其是在源文档编写得不好并且要翻译成多种语言时,效率提升更为明显。进行预编辑,目标语言越多,节省时间越多。 预编辑还与节省成本直接相关,这是语言服务提供商和客户都希望节约成本。 更重要的是,预编辑可以减轻翻译所受的困扰,因为预编辑可以使他们免于纠正机器翻译输出中经常出现的一些重复性错误。 最后,最终用户体验也将得到提升。 因为本地化内容变好,意味着客户体验也会得到提升。 适合预编辑的情况 在使用几种目标语言的大型项目中,可能需要考虑进行预编辑,因为语言数量越多,节省的时间和成本越多。 但是,预编辑并不不能解决所有的机器翻译错误。在某些情况下,预编辑并非绝对必要,或者根本不能带来好的投资回报率。如果源文本的质量已经很高,并且机器翻译引擎已经在特定领域的资源上进行了培训,那么预编辑带来的益处可能很有限,并且花费在预编辑上的时间也不会因缩短后发布时间而得到补偿。 其它要考虑的问题 有关预编辑的讨论还涉及一个问题,即原文的作者是否应该意识到,他/她的文章将来可能在某些时候翻译其作品甚至进行机器翻译,并据此调整其写作方式。 一方面,如果作者考虑到了,就可能减少对预编辑的需求,提高翻译质量和时间效率。 但另一方面,它可能会限制作者的创造性表达,并影响目标语言原文的目标读者的体验。