Machine translation with human post-editing

Post-editing means revising a translation generated by a machine translation engine with the aim of improving its output and, obviously, this also makes the cost of a translation cheaper.

It is essential that the source text is as well written as possible. Therefore, it is often necessary to edit the original text before it undergoes translation. Obtaining a source text that is clear, simple, free of typos and unambiguous is essential for achieving the best results.

Once we have a good base text, it undergoes machine translation that uses a translation engine. Documents are never displayed or posted on a publicly accessible platform.

As professional translators, we add this engine to our translation memories and termbases, which makes the process much more rewarding and can give surprisingly good results.

The better written the original document, the better the final text will be.

The answer is no. Post-editing is useful for contents with a highly structured language, so it usually works well for technical texts, which are characterised by their objectivity and do not contain double entendres or wordplays.

Therefore, the quality of machine translation must be assessed in every case.

It is fundamental to start out with a good source text. The clearer, the more accurate and grammatically correct a text is, the better the result of machine translation and the lighter the post-editing will be. As a result, productivity will increase and rates will drop.

In our case, totally. We resort to private, customised engines designed specifically for professional translators. None of the documents entrusted to us are ever displayed or posted on a publicly accessible platform.

Depending of the degree of revision, we talk about light or full post-editing.

As its name suggests, light post-editing is when we introduce as few changes as possible. The objective is that the text is understandable, but without improving style or flow.

Full post-editing, on the other hand, requires deeper revision. The aim is not just to make a text readable, but to manage to make it coherent in terms of a client’s context, style and specific guidelines. The outcome is a text that flows well and is completely coherent, almost as if it had been directly written in the target language.

No, it’s not the same. Mistakes that a machine makes have nothing to do with those that humans make. Therefore, in doing post-editing work we mainly concentrate on revising inconsistencies in terminology (machines don’t think!), omissions, flawed translations, unsuitable registers, etc.

The machine does part of the work, but then humans (and professionals!) step in to bring coherence and ensure all of a client’s specific guidelines have been followed.

Clearly, if it works well, this process can reduce the rate charged for a translation.
In the case of light post-editing, for instance, the cost can drop by between 30% and 40% depending on the language, how well the original text is written, etc.
In full post-editing, the cost can come down by up to 25%.
However, in order to give a final rate it is essential that the source text be analysed beforehand.