Hanna Martikainen conducts PhD research at Paris Diderot University on the French-language translation of Cochrane Systematic Review abstracts. Having previously worked as a translation coordinator and post-editor for Cochrane France, she now teaches a class in post-editing medical LSP at the Department of Cross-Cultural Studies and Applied Languages (EILA) at Paris Diderot University. During the past academic year (2016-2017), second year Master’s students in specialized translation have had a unique opportunity to contribute to the French translation project by post-editing machine-translated Cochrane abstracts and Plain Language Summaries. Now that the project has been renewed for the coming academic year, it seems like a good time to reflect on the first experience, presented here in the students’ own words.
Marie, Angèle, Margarita, Pierre, and Pauline described the profiles and general background of students in their second year Master’s degree in specialized translation:
“We are a group of 20 students from the ILTS (Industrie de la Langue et Traduction Spécialisée) Master’s degree at Paris Diderot University This course is open to students with prior translation experience in which is the case for most of us, but also to students who have technical experience and who speak at least two languages. Each of us does a work placement in different companies where we translate documents from fields including medical, legal, transport, entertainment, automotive, marketing, finance, tourism, and engineering.
Before starting post-editing with Smartling, we already had experience with Systranlinks, but none of our courses had trained us to do medical translation. Regarding CAT tools, we had previously learnt to use Trados Studio and Memo Q (Adriatic). As part of our training, we carry out different translation projects related to real work demands and Cochrane is one of them. Each one of us had the opportunity to post-edit five medical texts for Cochrane. It was a great experience because we discovered a new field and a new way to work with languages.”
Other students discussed the specificities of Machine Translation and Post-Editing. Vera, Marie, Lionelle, Marion, and Justine drew on their own experience to formulate some useful tips for beginner post-editors:
"Post-editing consists of editing automatic translation produced by a machine. You may already have used machine translation such as Google Translate. Post-editing will undoubtedly play an important part in the future of translation, since it will help improve productivity and reduce costs.
There are several differences between human and machine translation: first, style is usually not a priority when post-editing, the most important thing is to convey information. Machine translation is not suitable for every type of text; indeed the output is better with technical documents (such as medical and industrial content). The more repetitive the content, the better it is for the machine.
Automatic translation by the domain-specific engine is mostly terminologically correct, but often syntactically lacking. During the post-editing process, we often had to rephrase or rearrange long and complex sentences (for example, sentences with complex noun groups were often wrong). Other kinds of frequent mistakes include: terms not being translated, mistranslations, grammatical errors, and inconsistencies."
Their advice to beginners who would like to post-edit are:
- Try and read the whole source text before post-editing.
- Do not focus on style more than necessary.
- Do not blindly rely on the platform’s suggestions.
- Keep in mind that the final text has to be of publishable quality.
- Enjoy it!
Another group of students wrote about the specificities of medical English and their importance when translating. Here are some insights from Nadjet, Mélanie, Pauline, and Lucie:
“Regarding terminology in medical content, we noticed that English terms do not necessarily belong to languages for specific purposes (LSP), they often seem very simplified. In French, it is quite the opposite since specialized terminology of Latin origin is almost the only one to be used in both LSP and general language, including in hospitalization reports for patients, such as:
- baby → ‘foetus’
- stroke → ‘accident vasculaire cérébral (AVC)’
- back pain → ‘dorsalgie’
- blood disorder → ‘affection hématologique’
Also, in English, noun groups containing various modifiers of a head are packed, whereas in French, they often need to be unpacked, e.g.:
- TLR2-induced protein → ‘protéine induite par les récepteurs TLR2’
- main study visit → ‘visite au cours de l'étude principale’
- head and neck cancer → ‘cancer des voies aérodigestives supérieures’
- still birth → ‘naissance d'un enfant mort-né’
Finally, the use of passive sentences in English is very common, while we are instructed to avoid them in French as much as possible. However, translating medical content in French implies the use of a lot of passive structures that are very difficult to avoid.”
Last but not least, Joanna, Maya, Camille, Bruno, and Mickaël discussed how their pre-conceived expectations compared to the reality of post-editing:
“We were expecting a system similar to Google translation in its early years – huge terminology issues, not to mention typography and a catastrophic writing style. Actually, it wasn’t as bad as we thought it would be. Terminology was generally quite on point, as was the syntax, though there were several cases where we had to make the sentences more fluid and less 'robotic'.
Despite spell checking and grammatical software, the machine made more mistakes related to language than meaning. The longer the sentences were, the more mistakes occurred. It was sometimes difficult to refrain from rewriting the entire sentence when it was too long or when the syntax was wrong.
We also thought it would be necessary to have medicine-related knowledge; that the various diseases or health problems described in the text were not going to be explained, since it would have been written for people already familiar with this research. In reality, the terminology was most often correct, and the author described the research subject very precisely in the introduction. The first paragraphs gave us a good grasp of the subject. We simply needed to do some extra research to make sure the terminology was accurate and we had correctly understood the process of thought. Of course the machine version is not entirely reliable but it was far more efficient than we expected.
One of the main difficulties was to work as a post-editor and not as a translator. We had to learn to work fast, and accept a sentence that is not perfect. The use of automatic translation takes off some of the stimulation that traditional translation gives you, since the work is already done - at least partly - and you have to keep the proposed structured in order for the process to be time-effective.”
In conclusion, post-editing machine-translated Cochrane texts gave the students a hands-on introduction not only to this recent and booming domain of translation industries, but also to the specificities of medical translation. Moreover, the experience was a professionalizing one, given that the student post-editions were all published on Cochrane websites with acknowledgment given to post-editors.