Automatic translation is not the same as Google Translate

 Is it true that machine translation and Google Translate are not the same thing?

Despite what is commonly believed in the localization industry, machine translation is not free from the stigma that Google Translate has created. Localization experts, when dealing with a client interested in machine translation, face the problem that many people tend to unify the two concepts and refer to the two terms indistinctly as if they were the same. Although, on the face of it, we shouldn’t be surprised, given the massive presence of Google in our lives. Which leads us to the following reasoning: “well… if this is the best Google can do, why am I going to keep looking…”

Google’s automatic translation

Here it is important to clear up misconceptions about Google Translate and machine translation in general. Okay, that’s right: Google Translate is an example of an automatic translation system. And to be more precise, it is an automatic translation system based on statistics. But it wouldn’t be fair to rule out any kind of automatic translation either if, when translating the content of our website with Google Translate (for example, into French) and presenting it to a French-speaking audience without a native speaker having reviewed the translations, we were to come across a horde of customers and suppliers who were furious that you didn’t know which translation agency we had commissioned to do the work. But what can you expect from a statistically based machine translation system that is not focused on any particular sector, but has been created using publicly available translation memories? And if you have ever published any kind of content by translating it with Google Translate and without hiring any kind of review, correction or human editing, you have no reason to complain, nor to blame the machine translation for a mistake you could have easily avoided.

Machine translation has great potential. It’s actually possible to set up an automatic translation system to match your company’s content, which can’t be done with Google Translate. This is the main difference between Google Translate and a configurable automatic translation system, of the many on the market. By training the machine translation system using bilingual files (usually translation memories, glossaries and dictionaries), it is possible to provide yourself with a fairly good localization tool to cover your company’s translation needs, reduce the time to publish content and also save on production and localization costs in general. This in turn will allow us to localize and translate more content, which we could not afford before because it was not a priority, and to make the most of our localization budget.

This machine translation thing sounds good, doesn’t it?

At this point, let’s not get carried away with euphoria either. Realistic expectations should be created and the issue should be approached with common sense. Machine translation can provide very good literal translations depending on the type of content. But copy content used in marketing and advertising need a creative touch that, of course, is not provided by machine translation, and therefore we should probably rule out machine translation for this type of content. However, for user manuals, help files and other technical documents with highly repetitive content and minor syntactical variations, machine translation may be an option to consider.

Setting up a machine translation system also requires the involvement of technical staff and a process-oriented structure, something that can only be achieved with a team of professional translators and linguists. When it is done well there is no doubt that it is an excellent working tool for translators. Let’s be clear about this: just as translation memories are an indispensable technological tool for translators in their workflow today, machine translation should be given the same consideration. In the vast majority of cases it is necessary for a human translator to review and correct all content to ensure that it can be published. And when all this revised, corrected and edited content is put back into the system for a long time, the quality of the machine translation will undoubtedly improve. This will mean that our translators and linguists will have to make fewer corrections in the post-editing phase and therefore translate content faster and at a lower cost, thus increasing productivity and resolving the eternal conflict between Google Translate and machine translation.