How to optimise texts for machine translation: tips from the experts

machine translation

Would you like to use machine translation (MT), to save time, reduce costs and get a translation that suits your needs?
Did you know that you can optimise texts for machine translation so as to get the best results?

Assessing whether your text is suitable for machine translation is our job.
But writing text that’s already optimized for translation is something you can do, too.

We would like to share with you some useful tips for drafting copy that you can use not only to prepare your texts for machine translation, but for all your communication materials, including those intended for human translation. Clear, correct and unambiguous text is appreciated not only by machines, but also by your human readers!

As we know, machine translation is based on the use of artificial intelligence.
Artificial intelligence cannot properly handle some types of texts, especially those in which cultural, emotional and stylistic factors are important, such as texts for advertising and marketing purposes. However, even in technical texts the machine can run into difficulties if it encounters ambiguities, long and complex sentences, syntax or spelling errors.

In some cases, it may be useful to work on the text to be translated, making a series of corrections and changes before “feeding” it to the machine translation engine. This will help the machine understand it better and deliver a better result, which will then be the starting point for any post-editing.

This step can be crucial to the success of a machine translation project. In technical terms, this is called pre-editing. While this is a process that should be carried out by experts, there are a few things you can do to make your texts more suitable for machine translation:

  • Make sure the text is in editable format (e.g. .docx) and not in .pdf or image format. The translation software must be able to read the content of the document and edit it, which is not possible with .pdf files.
  • Avoid using poorly understandable text. Texts that are already a machine translation in themselves or that have been written by a non-native speaker will most likely generate a poor quality and therefore unusable machine translation. The post-editor would be forced to rewrite or reformulate almost everything, as if it were a translation from scratch. The time and economic advantage of machine translation would therefore be lost.
  • Check the text layout. If you perform an OCR conversion of the original document from .pdf, try to optimise it to correct errors and fix the layout. For example, use automatic bulleted and numbered lists instead of manual ones, insert indents and tabs correctly, and eliminate line breaks within sentences.
  • Check the text for typos, spelling or grammatical errors. Ensure that punctuation is consistent and correct, and capitalisation is used appropriately. The machine is not able to interpret these types of errors and correct them as the human eye can, often unconsciously.
  • Use short, simple sentences (preferably no more than 25 words), with few parentheses and few subordinate clauses, following a clear word order. Avoid double negatives. If necessary, divide the concepts in each long sentence into two or more short sentences. The longer and more complex the sentences are, the more the machine struggles to translate them correctly.
  • Avoid abbreviations: only the most common ones are recognised by machine translation engines.
  • Use capital letters only for terms that are proper nouns.
  • Check that the source text has consistent terminology. Avoid synonyms and variations of a single term. For example, if you talk about “key” on page 3, it cannot become a “button” on page 15 and then a “pushbutton” on page 20. Also avoid terms that may have a different meaning in another context. Does “plant” mean a factory, or a tree? Better to use an unambiguous term. In fact, while it is easy for a human to interpret a word based on the context of the entire document, the machine translates sentence by sentence and can get confused if it does not find enough context within the sentence itself.

What advantages does pre-editing offer you for machine translation?

  • If pre-editing is done correctly, it can improve the result of machine translation and reduce post-editing time.
  • This time saving is multiplied when working on very large and/or multilingual projects: by improving the source text upstream, post-editing in each of the target languages will be quicker and easier.
  • Finally, you can apply these tips to all the texts you write: even your human readers will appreciate a clear, correct and unambiguous text!

We hope you find these tips useful when writing or preparing texts for machine translation. Remember that you can always rely on us: interlanguage experts will know how and when to apply pre-editing techniques to your texts.

Are you interested in this service? Contact us for a quote!

How to optimise texts for machine translation: tips from the experts ultima modifica: 2021-04-20T15:05:12+02:00 da