Post-editing a chapter of a specialized textbook into 7 languages: importance of terminological proximity with English for productivity

Access to textbooks in one’s own language, in parallel with the original version in the instructional language, is known to be quite helpful for foreign students studying abroad. Cooperative post-editing (PE) of specialized textbook pretranslations by the foreign students themselves is a good way to produce the ”target” versions, if the students find it rewarding, and not too time-consuming, that is, no longer than about 15-20 minutes per standard page (of 1400 characters or 250 words). In the experiment reported here, PE has been performed on a whole chapter of 4420 words (in English), or about 18 standard pages, into 7 languages (Portuguese, Japanese, Russian, Spanish, Bengali, Hindi, Marathi), native tongues of the participants. Average PE time has been measured, and when possible correlated with primary PE time (the time spent in editing a MT pre-translation in the PE text area). When terms are cognates of English terms (as in French, Spanish, Portuguese, and even Russian or Japanese), because neologisms are directly borrowed from English, or built using similar roots (often Latin or Greek) and similar word formation mechanisms (composition, affixation of special prefixes and suffixes), target terms can be ”guessed” and PE time is in the order of 15 minutes per page, even if the target language is considered ”distant” from English. On the other hand, PE times increase by a factor of 3 to 5 when the target language is terminologically distant from English. We found that ∗[email protected] †[email protected] ‡[email protected] is the case of Hindi, Bengali and Marathi, and probably of all Indic languages. Previous experiments seem to have missed that important point, because they were performed on too short texts (often, only a few paragraphs), and on ”easy” pairs like English-French. A consequence is that, for terminologically distant language pairs, one should begin by separately collecting, or if necessary coining, the terms in the target languages.