Machine Translation (MT) is a fairly polemic subject. On one end you have the popularized, utopian view of completely automated translation, with free online tools such as Google Translate. On the other, you have the language purist, who holds in their heart the idea that human translation will never be undone by the machine.
In practice, you can see it both ways. Who can deny the ease with which one can get the gist of a French hotel’s website with a quick click of an online tool? Sure, the translation may not sound polished, but you can figure out what time check-out is, and if they have free WiFi.
But what about when industrial machinery operating instructions are on the line? What of reference databases and technical training manuals? And would you really trust your ad campaign’s creative copy to the microsecond magic of Google Translate?
MT: Cheap vs. Appropriate
MT is vastly more complex as it pertains to businesses facing the reality of securing profit overseas via translation and localization. Is MT a tool for driving down costs and improving translation speed? It can be. Depending on the project, you can realize 80% of the quality for 50% of the cost. To make the resulting translation cost-effective and satisfactory, though, there are multiple factors in play — something we’ve learned through hands-on experience with our clients as the MT field has evolved.
Our thinking at Acclaro is that for projects in the next 15 years, 100% human translation will being reserved for the most high-context, high-value content such as customer-focused pre-sale material, and MT will take on a much larger role for companies that have very high volume projects on the low end of the ROI scale, such as massive backend support documentation or internal documents that require only “draft” quality.
Should my business use MT or not?
There are three primary criteria we look to when making that evaluation of whether to use machine translation or not.
First, project size is a major driver in the decision to use MT on a project. Good candidates tend to have one million words or more to translate. Without diving too deeply into the history of rule-based versus statistical MT and the hybrid systems which integrate aspects of both, there are up-front machine training investments which make shorter-run projects less eligible. Time is spent tailoring the authoring environment, preparing the machine for what to expect, and deciding on use cases based on the subject matter and industry jargon.
Second to project volume is the complexity of the source content. Complex, multi-clause sentences are difficult in MT, as are slang phrases and idiomatic expressions. The creative voice tends to get mangled in MT. This shifts the focus of many MT projects back to more technical content such as service manuals and support databases.
Third, target language can be a factor in assessing whether MT is a viable option for a major translation project. In our experience, many Western European languages tend to do quite well with MT, as well as Russian. Chinese and Japanese have proven to be more difficult. So when you’re weighing your international strategy and the role MT may play in it, knowing where you want to go matters.
Getting your business ready for MT
Global businesses, and fast growing businesses that have potential to expand into the global market, must prepare now for the future benefits of machine translation. Failing to think ahead can result in exponential costs down the road.
Forward-thinking businesses can start training technical writers to use simplified sentences, an active voice, and consistent core terminology. Defining a style guide with a controlled vocabulary is also a sound tactic for reducing expenses down the line.
While MT will continue to dominate technical, high-volume projects in the immediate term, it will be interesting to see where it goes next. It is a rapidly evolving field, leveraging advances in both computing and programming power to expand even further into areas of language translation.
With the explosion of user-generated content and plays by many social networks to expand their content holdings (LinkedIn’s recent aggressive effort to become a major business-oriented content hub comes to mind), there is certainly a need for increased cost-effective translation bandwidth out there.
What lies beyond the horizon will be exciting. For now, when a well-meaning associate worries for the future of my business, I’m happy to say we use a range of tools. Not the least of which is that great organic computer: the human linguist’s brain.
