Oct 16

Which quality do I want? ... Serious?

How goodshould a translation be? Are translation companies overdelivering? Are translation buyers underbuying? The topic of fit-for-purpose translation quality is one of the most hotly debated in translation forums and events today.

Free. Budget. Best Price. Premium. Expert. Quality-assured. Budget Professional. Standard. Business. Ultra. Pro. Reference Quality. Reading Quality. Publication Quality...

These are only a few examples of translation quality levels that a few minutes of scanning through translation company websites will discover.

Didn't we hear somewhere that a professional translator can only deliver one type of translation - professional translation? True. But what an individual may not be able to deliver, a process can.

But let's start at the beginning. What is all this talking about differentiated quality levels?

Up to very recently, any discussion about translation quality invariably was around how to achieve the best possible quality, and how one could measure it.

With the growing success of machine translation, a number of companies saw an opportunity to extend the scope of translation to a vast area of content that had not yet been covered by the translation industry: huge corpora of reference material, short lived on-line content, user-generated content. Why they had not been covered was plain to see: too much content, too short-lived, not enough budget, not enough professional translators to finish the job in time.

With the proliferation of higher quality statistical machine translation (the term higher is to be understood in comparison with earlier generations of MT engines), those non-covered areas became a potential business target.

Soon enough however, providers and buyers realized that pure machine translation output, while allowing basic understanding, often did not meet the user's objective. Humans had to edit the output to rule out misunderstandings or even make it pleasant to read. The post-editor was born, together with the understanding that one size does not fit all. Not all content types require a polished style or the same level of language accuracy and in some cases, a less-than-perfect understanding was acceptable too.

Smart purchasers immediately saw the potential for quick cost savings. Translation departments and other buyers of translation in the company however saw the spectre of translations that they would have to rework because the quality was much lower than anticipated.

Purchasers got what they wanted, and in many cases buyers got what they feared.

The lack of a standardized definition for any of the listed quality levels creates an expectation mismatch. What a provider calls "less than perfect", a client validator may categorize as "to be redone completely".

The perfect recipe for failure.

Does that mean that the concept of differentiated quality levels is wrong? By no means. One size indeed does not fit all. But without standard definitions any attempt by buyers and providers to agree on which quality level fits which need will inevitably be flawed.

It is also clear that not one single provider of buyer can provide such standard definitions. That is why we will be looking at the language industry associations to take the lead. EUATC and its national members certainly have a crucial role to play in this area!