Localization Trends That Never Arrived

Anyone in the industry for a few years will pick up on some of these "coming trends" that never seem to arrive:

* There will be an industry leader (think the Microsoft of translation)

Instead we have companies growing by being bought up and merging (think, "we are all out of ideas of how to add value and gain customers thus we can just buy other company's customers")

* All translation will be done online, for free (Wikipedia style) by the cloud

This is a misunderstanding of the most basic tenants of the art of translation—that not everyone who speaks two languages is instantly a translator.

* All human translation will just be editing of machine translation

Most professional linguists will not work for a cut rate to edit often substandard machine translation—"first we have to spend thousands on translation tools so we can get paid a lower rate for our work. Now MT means an even more fragmented rate."

* Maybe not said, but suspected: Translators will get smart, work directly with end clients and cut out the middleman (after all, programs like Trados include each linguist's email right in the TM)

This thought goes hand in hand with the dream that online systems will eliminate the need for the project manager, the SLV, etc. Linguists cannot work directly with clients because they cannot take the incessant (and sometimes crazy) demands of end clients--as well as the big MLVs and their frazzled intern project managers.

Of course our industry will change and is changing, but it really seems that over the years there has been an exuberance of declaring some new trend will fundamentally alter the business--and then we never see that trend arrive.

Magical systems to eliminate translators from translation

The rush to install time saving systems so that we can be more efficient and streamlined is, after all this time and evolution of the processes, still a dream that is not realized.

We work with many clients who seem to have a lot of faith in a technology based solutions. These promise to cut costs (mostly their personnel costs) with a magical system that will streamline the process, eliminate the need for an experienced localization industry project manager, and have translations inserted into an online platform that makes delivery to client automated and faster.

Truth is that no such fairy tale is happening. We are out here in the real world, working consistently to deliver high quality translations that will fully meet the end client's expectations--often within demanding delivery schedules. Most of the time we are working around such systems.

We are finding that the many language service providers are being sold into translation platforms and project management systems that raise expectations far beyond what is actually available out here in the real world.

Let us clear away the hype and try to see what is going down at the coal face where the real work is actually being produced.

Most translation work done is today, just as it always has been, by professional freelance translators working to support an ever growing industry. These professionals are working feverishly every day trying to rise to the challenges that are constantly thrown their way by the software and platform generating gurus who only seek to cash in on their work by promising translation buyers even greater returns through their "revolutionary" systems.

Then there is the rush to have a computer do it all. However, in reality, there is no escape from the simple fact that machine translation output is unacceptably poor. So now there are demands being put on professional linguists to convert such output into high quality target language text by a process now called PMTE (post machine translation editing). More and more we are seeing this being requested as if it were a legitimate localization industry process.

The founders of Admerix have been closely involved with the introduction and development of machine translation since 2004 when we attended seminars that promised super high-speed multi-language machine translation from any language to any target language. The boast was that this would be achieved within six months and it would have a quality standard of 96% that of human translation.

14 years on and that prediction seems to be just as unlikely as ever it was.

Of course, the offered payment for PMTE is a fraction of the normal translation rates. Translation buyers, charged by the hype of those with large investments in machine translation as well as some large LSPs who are looking to fatten their margins, are working hard to get professional linguists to take on full responsibility of final translation results at a fraction of the normal translation rate.

For now, most professional linguists are not interested, nor is it cost or time effective, for them to try to re-translate the mess that most machine translation produces.

Can you edit from Indonesian to Malay?

This comes up enough that I thought would we should mention it here: These languages are not similar enough to be able to edit from one to the other. Oddly enough, many linguists in both languages believe it can be done, but see what they say when you give them a project and ask them to do it!

There is even a Wiki page on this, see here.

The Realities of Editing

In the real world we still depend on expert linguists despite the lure of cheap crowd-sourced work or an automatic miracle.

We also know that even when the most qualified subject specialist is employed on a project, others who edit/proofread/check their work will attempt to take it apart, often for completely unjustified reasons, perhaps as a justification for their own employment as an editor. This is just a syndrome we have to account for in the very subjective realm of localization.

So even top experts are disputed by other experts. Augmented or MT does not really address, in any way, the actual quality of a translation and the subjective decisions that have to be made.

Where it does come into play, we suspect, is as another attempt to chip away at what translators are paid. First it was the translation tools like Trados which had as their first advantage that an employer could discount repetitions and not have to pay a full-word rate for each word translated.

Now exactly the same thing is happening with MT--linguists are asked to assume every word is essentially a repetition and then vouch for the work for a fraction of their per word rate (or even an hourly rate). It seems incredible that any professional linguist would willingly accept such a bargain.

This race to minimize or eliminate the cost of the human element in any field is always going to be a driving element in business. However, the editor and the subjective decision are likely to remain part of the industry for a long time to come.