Translation is when we reproduce the meaning of an original text (or other type of content) in the target language.
Modern-day translation is performed with the help of specialised tools, such as CAT software, MT engines and translation memory.
Translation is different from localisation in that it only focuses on the text, while localisation focuses on taking a whole product or service into a new target language. It is also different from transcreation in that translation retains the meaning of the original work, while transcreation involves creating text from scratch in the target language and context. These definitions sometimes overlap.
You will find detailed information on different types of translation (by content and purpose) in the [link] section of the Translation Essentials guide.
Translation editing is when we verify a translation with reference to the source content. The translation editor reads the content in both the source and the target language to identify errors, omissions and misinterpretations. They can also improve the style of a translation.
Translation proofreading is when the translated and edited text is checked for any remaining errors, misspelling, incidences of poor style etc. Proofreaders do not refer to the source content unless they notice some logical inconsistencies.
Historically, copyediting used to be performed on a manuscript, while proofreading was performed on print.
The above-mentioned elements are the traditional translation framework, and the three combined form the so-called TEP (translation-editing-proofreading) gold standards for high-quality translations. However, modern translation processes often involve machine translation and post-editing instead of human translation.
Machine translation (MT) is when translation is performed by a machine-translation engine instead of human translators.
Professional use of machine translation is different from the browser-based web app we all know from everyday use. The way we use it is via plug-ins within our translation software.
Translation professionals and translation agencies use different types of machine-translation engines – statistical, neural or mixed. Each type has different pros and cons. Some of the biggest MT providers are Google Translate, Amazon Translate, Microsoft Translator and, less known outside the language industry, DeepL Translate. Each MT engine has fields, types of content and language pairs that they are best at.
Machine translation, despite enormous progress, still needs thorough verification through post-editing.
Post-editing is when a human linguist reviews the output of a machine-translation engine.
Post-editing can be challenging as, unlike in old machine translation, the output of modern machine-translation engines feels relatively natural. Post-editors should not be fooled by smooth language and must ensure there are no omissions, additions or misinterpretations in addition to plain translation errors.
It can be assumed that machine translation and post-editing combined are comparable to human translation in terms of quality.
And, just as human translation, it requires further steps – like editing, proofreading and LSO.
Language sign-off (LSO) is a check of the text performed in its final visual form and context.
For a print or pdf publication, it requires assessment of the text after type-setting, with all illustrations, pagination etc. in place.
For an online banner ad, it means assessing text embedded in the planned location on a website.