As you may already know, post-editing takes place after a text has gone through machine translation (MT), as opposed to pre-editing, which consists in making adjustments on the source text to prepare it for an MT engine and avoid any ambiguity or mistranslation. Post-editing is different from the work of a translator, whose job is to translate the entire text, and it can be carried out at different levels (e.g. light post-editing or full post-editing, which are discussed on the blog post about the future of translation written by Lucie Otto), depending on the corrections you want to add to the output. There are many tools to help post-editors in their jobs, which we are now going to look at in detail.
- Machine translation quality estimation (MTQE)
Machine translation is making incredible progress today, but despite MT innovations, the quality of a text produced by MT is unpredictable. So, it is essential to be able to estimate the quality of your MT output in order to know how much time you will need to post-edit it.
Tools such as machine translation quality estimation, an AI-powered feature in the Memsource Editor, can provide quality scores for MT outputs. That way, you can estimate how much work is required for each specific segment. For instance, you can find the following scores:
- 100% MT: the MT output is perfect and probably no post-editing is required;
- 99% MT: the MT output is almost perfect, but there are a few formatting or punctuation issues;
- 75% MT: the quality of the MT output is high, but is worth post-editing;
- No score: when there is no score, it means that the quality of the MT output is very low. This MT output should not be published under any circumstances and should be used for reference only.
Quality scores also allow post-editors to know which segments should be prioritised: you should rather focus on segments with a low score than on those with a high score.
You can also find a similar tool in the SDLTrados interface called MTrans Post-Edit Booster. The goal of this plug-in is to detect any style-related errors frequently found in MT outputs. This is a very useful tool for post-editors who are tired of having to deal with the same type of error when post-editing MT outputs. Such tools are essential time-savers when you have a lot of work to do.
To quickly sum up, you can decide whether it is worth using MT outputs in your work, or not, with this kind of machine translation quality estimation tool.
- Terminology management systems and reference documents
In order to ensure consistency within the text, you should use all the material that your client gives you. Make sure to use the translation memories, term bases or glossaries provided: translation memories allow you to reuse segments that have already been translated, term bases give you terms that have been verified or approved by your client and glossaries are kinds of brief dictionaries with a collection of terms and their meanings. This step is very important, because your client might want to use specific terminology and syntax, and you should be aware of this to avoid any mistake.
Other types of documents to keep in mind are all the reference documents that your client has provided such as style guides or terms of reference. These documents should be your guiding principle to ensure consistency. They will give you essential information about what type of typography you should use, how to quote a report or a book, and how to write names of countries, numbers or footnotes for instance.
These documents are the basis for your work and they are to be kept up-to-date for future projects.
- Use of Quality Assurance Tools (QA Tools)
QA tools can also prove to be very useful. They can either be integrated to your software or standalone. QA tools allow you to identify issues in the original output and errors introduced during the post-editing process. Among the best-known standalone QA tools is Antidote (available in English and French): it is useful to correct every aspect of the language, such as spelling and grammar, typography and style and repetitions. Thanks to this software, which has smart filters, you will be able to correct any capitalisation mistake and wrong verb agreement, or to delete unwelcome commas and redundancies. Another major pro of Antidote is that you can read the integrated guides and dictionaries, which will help you with grammar rules, synonyms and collocations.
To sum up, QA tools are the very last line of defence in the translation and post-editing process, and their goal is to ensure the high quality of the final text.
To conclude, post-editors have a wide range of tools to use at their disposal, and their job is to make the most of it, to move from a poor MT output to a high-quality translation (which could have been done by a human translator).
Lola Bathion
M1 TSM 2021-2022
La Traduction de A à Z 