The case for and against direct update of TM's by translators is an interesting article at the Localization Best Practices blog.
Noting that all translation service providers are looking to reduce any and all costs that they possibly can, the post wonders whether the practice of multiple translators simultaneously working off of a central translation memory ("TM") is compatible with the blog's assertion of an industry best practice of refraining "from TM update until after linguistic QA, and if possible until after client review".
In a related discussion on LinkedIn, two main themes emerged:
- More and more, the old TEP (i.e., translate, edit, proofread) model is looking tired and outdated. In the age of turnaround time pressures and crowdsourcing translations, new approaches are needed.
- The type and format of the content being translated has a huge impact on the feasibility of this approach.
The good news is that an organization doesn't have to spend that kind of money. Simple process changes can achieve 80% of the same results, with minimal costs. At ForeignExchange, we utilize our "incremental leveraging" and "repetitions file" processes to achieve nearly all of the same benefits of a central TM.
And, going back to the article at Localization Best Practices, we have successfully managed dozens of translators working in parallel, all using one TM, with excellent quality and no central TM.
Was this post useful? Subscribe to Medical Translation Blog via email or RSS!
2 Comments:
Subscribe to:
Post Comments (Atom)
Outside this controlled environment, in 17 years of working with company TMs, the material I have received has ranged from somewhat useful to extremely bad or counterproductive (probably the product of dumping aligned documents without any human input). While it is possible that with a massive annual translation volume of many millions of words produced by extremely good translators working according to certain guidelines, you might end up with fairly decent TM, I have my doubts.
These remarks are not meant to disparage the idea, just to question whether human input can be easily minimized. I expect that the most promising results would come from a comprehensive approach to authoring, terminology maintenance and translation, an approach that requires a serious, and expensive, commitment to company documentation.