Most technical manuals are sent to translation after an update, extracting content from either an available, scanned printed doc or a locked pdf. The necessary OCR process generates an editable text containing just “optical” line breaks there, where no ones actually exist, as shown in the following example:
Check the pre-treatment
limiting value setting
The CAT tool identifies two segments here, setting up two translation units, respectively, as follows:
TU1,EN: Check the pre-treatment
TU2,EN: limiting value setting
We call this a “bad source”… when translating the sentences into a Romance language like Spanish, we cannnot avoid inverting the genitive sequence – otherwise an invalid string will be created.
TU1,ES: Controle el ajuste del valor límite de pretratamiento
TU2,ES: [void segment!]
While TU1 results non-accurate, TU2 is empty – definitely both TUs are discarded for further usage! It is not different with a German source like “Produkt- / Beschreibung” (product description) as the following automatic segmentation shows:
If the decision is taken to write necessarily content on both target segments, then it leaves us no other chance than
TU2,ES: del Producto
where the resulting TM assignment (DE>ES), Produkt-=Descripción, Beschreibung=del Producto is again discardable. Joining segments in the CAT environment is not always allowed either by the CAT tool itself, or by the client.
Still all this can be worse
Longer German sentences are the worst case when trusting on Abbyy & Co. After being artificially segmented into two or more chunks, their inverse sentence construction -placing e.g. the main verb(s) at the end- just generates a set of strings without parallel regarding any Spanish construct.
According with our experience, a few number of customers and intermediary LSPs than expected are ready to understand this issue from the translator’s perspective as well as its negative impact on final translation quality and further translation memory usage; in our daily work we have decided to warn customers about this and started to re-process the supplied source docs joining all affected segments to longer, complete sentences, before sending such content to the CAT stage… if allowed. Unfortunately, some intermediary LSPs provide pre-processed bilingual source content (e.g. .ttx files) spoiled with the mentioned drawbacks, and on top of that, with the additional requisite to give back a high-quality TM and -frequently- without any possibility to review the final layout & artwork.
Having in mind parameters regarding the best product quality/working satisfaction compromise we have decided simply to decline accepting such assignments.
This blog post has been contributed by Alejandro and Elizabeth, they are German, English to Spanish principal translators at TranslationArtwork.com. If you need a professional translation service in any language, at TranslationArtwork.com your translations are in safe hands, believe me. As a translator you a welcome to easily subscribe for translation jobs here.