The vision of creating an effective, cheap, accessible, automated, and independent system for localization has long been a fundamental source of progress in the translation industry. One of the early ideas on how to go about this was the idea of a “community” translation model. Unfortunately, the idea proved too advanced for its time and was abandoned.
After a long time, global Multilingual vendors (MLVs) are now reinventing the Community Translation Model (CTM) in order to put it into praxis yet again. While past efforts to run this model have proved unsuccessful, this new model might possibly turn dreams into reality. In fact, a trial of this new model is currently underway. How should translation buyers look at this?
In this post we will look at what a “cloud community” is, how it once failed, and how it should work.
The CTM basics
The currently emerging CTM is essentially an agency-managed virtual (i.e. cloud-based) group of freelance translators “existing” within an agency’s translation management system. Translation orders are shared directly into the cloud and are processed on a first-claim-first-do basis. The delivery of completed tasks is quick and clean. The overall process aims to be fully automated, with human supervision at the ready.
A bit of CTM history
The first to experiment with the idea of CTM in praxis were not the MLVs, but big global software houses. This all took place at the end of 20th century, which was the peak of pioneering in localization processes and the launching of CAT tools. Among the best-known attempts was that of Apple. At the time, Apple’s MacOS environment held the position of the most adored and admired operating systems around the globe. Apple, as a big visionary, utilized this image and put together a global translation community of devoted Apple enthusiasts. The enthusiasm was so great that at first, translations were even performed for free. The resulting translations, however, were received much less enthusiastically. The localization process was simply not technologically mature enough at that time, and the resulting community translations were of such horrible quality that they could, arguably, have jeopardized the company’s perfectionist image. The unsatisfactory result of this attempt was long considered the nail in the coffin of CTM. Eventually, Apple became one of the most demanding translation buyers with its perfectly tuned traditional translation processes, and talk of CTM died down.
However, although other visionary ideas were also wrecked at the beginning, they came back like a wrecking ball when the time was right. Let´s remember the Apple Newton, the true ideological predecessor of today’s smartphones and tablets. Having made its name as one of history´s biggest tech-flops, Newton as an idea was successfully reinvented and now forms the basis of key features in the iPad and iPhones.
Could this perhaps be the case with CTM as well?
Parallels between CTM and Machine Translation (MT)
The current relation between CTM and existing human translation resources is the same as that between MT and the existing translation memories. CTM could not exist without trained experienced translators, just as the MT engine cannot produce any reasonable translation without having been thoroughly trained using existing translation memories.
What quality can be expected
CTM is like MT even in terms of quality, at least to some degree. Both have the potential to produce imprecision and varying quality of translation. Further, the extent of these imprecisions and differences in quality may in both cases depend on the language to which the CTM or MT is applied. Finally, just as with the MT engine, the quality of CTM output varies with the size and quality of the available translation memories and the subject area.
What the buyer should be aware of:
Even though CTM might increase efficiency in some ways, it is not a miracle cure that fits all sizes. In and of itself the model has many limitations, mainly in terms of the resulting quality. As the employment of CTM is driven by a thrifty, economical approach, so its utilization should be well considered from many points of view. (We have heard from many an unhappy freelance translator about the financially unfair offers they have received from existing CTMs run by well-established agencies.)
Due to the “first-claim-first-do” project assignment in the CTM model, CTM translators can never attain the level of expertise maintained by the traditional project assignment. While in CTM texts from the same subject area, or even the same company, may be translated by numerous translators with varying qualifications and levels of experience, the traditional agency arrangement ensures the continuous assignment of projects within a particular subject area to dedicated translators.
CTM should be used with linguistically clear content that uses simple text. This means the source text should be well-structured and easy to read. Good uses for CTM are, e.g., user documentation for established operating systems or management systems like MS Windows. Also, simple e-marketing texts like those of Amazon or Alibaba are good targets. SAP or new products that are at the beginning of their global business journey would be less suitable for CTM. Where CTM falls completely short are in Life Sciences, legal documentation, and creative marketing translations. If MT is used in conjunction with CTM, care should be taken to use an engine that has been trained on really large and linguistically consistent TMs from the past – because if the MT engine is insufficiently trained, post-editing by the CTM will be prone to errors (by its nature).
What the buyer should understand:
Establishing a good CTM is a challenging task, and not every big agency is prepared to set it up properly. Buyers should therefore carefully consider:
How does the agency select and qualify its resources (i.e. translators) in their cloud community? How do they measure their quality? How is their continuous improvement monitored?
How powerful, in terms of productivity and the covering of peaks, is the agency’s CTM really?
What is the level of task automation (the reality of the cloud portals and CAT tools used by the agency)?
What kind of linguistic support do the CTM translators get, and what is the accessibility of that support – i.e. glossaries, terminology support, style guides, training, etc.? How convenient is it to access this support – e.g., does a too-complex authentication discourage translators from accessing the glossary?
What is the process for reviewing translations – are there systematic quality checks performed?
What is the availability and what type of work are experienced reviewers assigned to – do they do QA on randomly selected samples?
For big languages such as FIGS, CTM usually attains a higher level of quality than for minor languages, so choosing CTM might be less appropriate for, say, east-European languages.
In Conclusion…
Even though the current CTM is perfectly set up, it is obvious that it is not suitable for everything. It is certainly not going to be the best choice for complex and sophisticated content where information is layered and could encompass more meanings. It could also underperform when applied to new products, where proper linguistic setup (rules, style guides, terminology, TMs) are not yet established. It is also not suitable, e.g., for medical devices, where even a simple imperfection could result in death. It can, however, serve well for efficient translations of simple and large texts which are continuously updated, where terminological and stylistic consistency is less important, and possible random translation mistakes would not cause any kind of damage to the product image or the user. In any case, we look forward to seeing how this attempt at a new Community Translation Model will progress.
In the meantime, we are here to support you every step of the way with the traditional high-quality services you can always count on. Learn more at: ABOUT exe’s LOCALIZATION BRANCH | exe
Team exe