The Good and the Bad of Automatic Translation
- 11 Feb, 2021
One of the most frequently asked questions we receive is: "Do you sell automatic translators?"
The simple answer is “No.”
The clarification for that answer will require us to explain why automatic translation is not only a not-quite-there technology, but also why it might be absolutely detrimental for your organization, whether it be a school, a house of worship, a corporation or any setting where getting your meaning across correctly and efficiently is paramount.
Let's get started!
Automatic Translation: Is it There Yet?
Definitely not! While many companies such as Google and Microsoft are leveraging the power of artificial neural networks, supercomputers and state-of-the-art machine learning technologies, these systems still pale when compared to even some of the least qualified humans.
Some companies irresponsibly declare that this type of software will “Turn your PC into a full-fledged translation center”; “This is the end of human interpreters; “You now have a ‘Star Trek translator in your pocket,” or other hyperbolic statements that have well-meaning people throwing their money away on inefficient – or sometimes outright nonfunctional – software.
As we have corroborated over our years of experience in the translation and simultaneous interpretation craft, marketing can only get you so far: when results are allowed to speak for themselves, the best and most expensive machine translation systems have extremely poor success rates.
Why is it So Difficult for Machines to Understand Language?
Many of our innate abilities are astonishingly difficult to transfer to machines, as we still know very little about what makes us tick – every one of us has language abilities which exceed even the most gargantuan supercomputer existing today. Let’s list some of the most frequent issues that baffle machine translation experts.
Context and disambiguation: computers, however complex they are, are binary systems that produce an output from a certain input. They have an exceedingly hard time attempting to “understand” which meaning to use when a word has multiple meanings. They can also bumble spectacularly because many of the cues a human uses to understand meaning come from context (environmental, gestural, by knowledge of the speaker, etc.) and not just from sound.
Machine translation systems have difficulties understanding when a word should be translated in isolation, and when it constitutes part of a larger meaning.
Non-standard language (which is most of our spoken word) presents another significant challenge, as all of these systems “learn” by integrating many samples of standard, formal language, and are completely at a loss when finding any sort of colloquialism, regional dialect or vernacular that doesn’t conform to extremely stringent and narrow definitions.
Idiomatic expressions: as an extension of the previous issues, all languages have phrases that cannot be translated arbitrarily, as their meaning is not literal; “It’s raining cats and dogs” may make perfect sense in English, but it may be harder to find a 1/1 correlation in Filipino or Japanese, hence even though some automatic translation systems can handle some idiomatic expressions, more often than not they mistranslate them, place them out of context or provide a literal translation which only exacerbates the “strangeness” inherent to any text not produced by a human.
What these difficulties usually amount to is an unusable, ineffective or straight-out unintelligible output. Understanding how and what to translate and interpret requires constant study by the translator/interpreter, in many cases attaining intimate knowledge of the subject at hand.
Is Automatic Translation All Bad?
Certainly not! It can get you out of a jam if you need to travel abroad and need assistance with reading a menu or understanding signage. Google Translate (and their exclusive phone earbud, the Google Pixel) can provide an OK tool if you have a smartphone and slowly and deliberately speak into it to ask for directions or attempt to carry a general conversation with a stranger who does not speak your language, but these applications quickly show their limits when encountering more complex speech or are used in any kind of serious setting.
Also, it can be a boon when reading a website in a foreign language or attempting to send an email in a different language. Edits, though, are almost always a necessity.
Why is a Translation/Interpretation Broadcast System still Necessary?
Thankfully, this might be the simplest answer yet! Simultaneous interpretation systems (FM, IR, Wi-Fi, etc.) broadcast the interpreter’s voice from a transmitter into the receivers or transceivers carried by members of the audience. These systems will still be essential even if a human interpreter is replaced by a computer or another electronic device –you will connect the transmitter (such as the Enersound T-500) to the computer or electronic device and the translation audio will be picked up by the audience!
We hope to have provided persuasive reasons why automatic translation is not yet anywhere near the necessary level to entrust it with your high-priority meetings, church services or classroom activities, but also compelling reasons why a simultaneous interpretation system is always money in the bank when you need an effective solution for all of your translation/interpretation needs!