Reporting on the Report: When Mainstream Media Gets Translation Tech Wrong (Almost Always)

Reporting on the Report: When Mainstream Media Gets Translation Tech Wrong (Almost Always)

Reporting on the Report: When Mainstream Media Gets Translation Tech Wrong (Almost Always)

As I scrolled through the day’s news on my phone the other day, one story caught my eye. It was about yet another “instant translator”. I cringed as I clicked through to the story, expecting the worst. I wasnt disappointed. There was much ooh-ing and aah-ing about the next big thing in human communication and yet at the core it was still good old machine translation.

Granted, we dont expect anyone outside the language services industry to know the finer details of translation technologies. But why is there such credulity about devices that promise to translate from “any language to any language”?

If youre a reporter and are about to write a story on the next big thing in translation tech, take a moment to read this post and save yourself the trouble (and save us the cringing).

1. Machine translation has been around for half a century

Instant translation is nothing but machine translation (MT) by a catchy name, and researchers first started working on it in the 1950s. It’s true that the technology has improved vastly since the ’80s and ’90s, with computing power becoming more effective and less expensive. Still, be cautious with words such as “breakthrough” and “cutting edge” when referring to MT.

Given that this technology has been around for so many years, you should also know that there’s considerable literature around it, and to understand a true breakthrough in MT development, you would really need to be in sync with all of the previous work, or get the opinion of an MT expert.

Some MT engines are open-source. Do check with the developers of new devices about how much of their tech they actually developed themselves.

2. And, human translation (and interpretation) has been around for thousands of years

The hype around Babelfish-like products is understandable: after all, wouldnt it be fascinating if humanity had a device to understand each other irrespective of which language they spoke? But understand this: that device simply doesnt exist. Even the most advanced interpreting devices (yes, interpreting, and we’ll come to that soon) do not cover even a fraction of the languages spoken by human beings, and for the languages that they do cover, there is still an error rate.

Compare this to the high quality translation and interpreting done by professionals that goes unnoticed. That’s another problem with mainstream media's coverage of translation: it only makes news when there are gaffes and, more often than not, MT is responsible for those embarrassments.

3. Repeat after me: all spoken translation is interpretation

Every time you equate translation with interpretation, you do a huge disservice to the interpreting profession. Interpreting is a fine art, and anyone in doubt only needs to sit through one simultaneous interpreting session. When a device claims to help two persons speaking different languages communicate, it’s technically an interpreting device, though of course, MT is happening in the background. This differentiation is important to make: interpreting is a more difficult process than translating, and hence, needs the discretionary capabilities of human beings more than written text would.

4. Understand that “ad hoc” needs are different from more structured, quality demands for translation

Take a moment to understand the kind of demand that may exist for these instant interpretation products. They might have a huge appeal among tourists wanting to know directions, make small talk, or find out bits and pieces of information. But how many of them would buy a device that’s priced above $200, which in most cases, they would use only for a niche purpose such as travelling? Time will tell.

But apart from such ad-hoc demands for translation in limited settings, there’s a 40 billion-dollar industry that thrives on continuous, consistent, and high quality translation demand. And most of that demand is met by human beings, word for painstaking word.

You should also know that MT engines learn from previous translations, which of course are done by human beings.

5. Translation tech doesnt start and end with MT

Machine translation is just one type of translation technology, yet it has unfairly appropriated all of it, at least in popular imagination. As human-intensive the translation process may be, there exist many, many tools that have automated processes and made life easier and faster for translators, project managers, and clients. No professional translator would survive a day without translation memory, terminology management software, other computer-aided translation (CAT) tools that enable collaboration, review, and feedback, and translation portals that automate file handling.

Of course, your audience may have little use for the nuances of these tools, but it still pays for you to stay informed.

6. How many is “all languages”?

The world currently speaks about 6,000 to 7,000 languages, and we need at least 100 languages to reach 99% of the worlds population. So dont swallow the PR spiel; instead, ask: How many languages exactly do you mean by all languages?

Translation and interpretation professionals help the world go round. Get up close with their fascinating world of words and you will find many a positive story about how the human touch has bridged gaps, removed barriers, and connected people. After all, this is one sector that has continued to grow regardless of business cycles. There is a lot happening with translation tech, too. Just don’t get swept up in the hype.