The Real Babel Fish?


New York-based company Waverly Labs have announced their new invention: The Pilot – the world’s first translation earpiece that translates conversations between people speaking different languages in real time. The company is planning to release versions for English, Spanish, French and Italian, while versions for East Asian, Hindi, Semitic, Arabic, Slavic and African languages will be available in the future. The device is now available for pre-sale but there isn’t a release date. In fact, there isn’t much information about the product on their website either.

Could this device be the real Babel Fish; the machine translation device in Douglas Adams’ The Hitchhiker’s Guide to the Galaxy? When placed in the ear, this small yellow fish allows the person to instantly understand anything said in any form of language. Is this just the stuff of science fiction, or can we really overcome the immense problems posed by machine translation?

We only need to use the free online translation websites such as Google translate or Yahoo! Babel Fish to see that machine translation is still in its infancy. There’s a joke doing the rounds of the internet that someone took the Biblical phrase, “The spirit is willing, but the flesh is weak” (Matthew 26:41) and used a free translator to translate it from English to Russian. When the same phrase was translated back from Russian to English it read, “The vodka is great, but the steak is bloody awful!”

When we use machine translation a lot is lost in translation. Here are just a few of the problems it faces.

Idioms. Idioms are metaphorical phrases such as on the house and at the drop of a hat. When idioms are translated literally into other languages they don’t make any sense. This problem is illustrated nicely in the Darmok episode of Star Trek in which the universal translator is used to translate the language of the Tamarin race. However, the language is based on metaphor from Tamarin folklore and it is completely obscure when translated directly into English.

Culture-specific words. Languages have cultural words for foods, objects, values and emotions for which there is no exact equivalent in other languages. If there’s no direct equivalent then translating a single word into another language may take many words in the target language, if the concept can be explained at all.

Ambiguity. Many common words have multiple meanings or senses (this is called polysemy in linguistics). For example, if someone says kind, are they talking about a “type” of something, or are they saying that someone is “good” in some way? Then there are unrelated words that sound the same (homophones). Are we talking about a pair as in two of something, or the fruit called a pear? We can ascertain the meaning of a word by its context, but machine translation isn’t always capable of differentiating meaning.

pic1If the translation device makes a mistake the results could be embarrassing, especially for business and diplomatic relations.

This was the case back in 2011 when Chinese premier Wen Jiabao made a high-profile visit to Malaysia.

There was a welcoming ceremony with a banner message written in Chinese and Malaysian. The Malaysian text said, “Official welcoming ceremony in conjunction with official visit of His Excellency Wen Jiabao to Malaysia.” However, the Chinese text read, “Official welcoming ceremony, with him Wen Jiabao His Excellency’s official visit Malaysia.”

It was later revealed that the Malay sentence was translated into Chinese using Google translate…

4435d1227458828-engrish-mild-naughty-jokes-image002Machine translation is constantly developing, but as you can see, there are many limitations and human translation is still far more accurate at circumventing these problems. Well, sometimes.

Of course, the above translation issues apply to human translation as well.

We need only look at a website such as Engrish to see that human translation can be a challenge too…


Posted in: Uncategorized
Karen Stollznow

About the Author:

Dr. Karen Stollznow is a linguist with a background in history and anthropology, and is the author of the best-selling book God Bless America. Her other titles include Hits & Mrs., Language Myths, Mysteries and Magic and Haunting America. Her forthcoming books include Missed Conceptions and Not What The Doctor Ordered.

3 Comments on "The Real Babel Fish?"

Trackback | Comments RSS Feed

  1. Shane says:

    Not to be too doom and gloom but the thing with wearable tech is the technical challenge isn’t in the hardware or the form factor, it’s in the software. They have been doing two years work to manufacture a physical item with streaming audio.

    The really large software challenges remain. Just recording language and converting it to Text is a large feat as anyone who has used Siri can tell you. Dragon are perhaps the rock stars in this space but still require hours of training to get it to 99%

    Opening that up to untrained voices is extremely difficult (especially if muffled outside) and we tend to only get a 90%+ accuracy with extremely clear commands. Actually travelling with this device expecting it to even transcribe accurately would be a huge world changing technical revolution.

    The second technical challenge is detecting languages. Chrome can quite easily detect (most of the time) a written language, but it uses clues from the characters in the print and usually has a large sample of text to scan over. Accurately predicting a language from sound waves is a whole unsolved problem, especially when it would need at least a few sentences (if not paragraphs) to feed the algorithm. This alone would be a massive technical leap.

    The last technical wonder is accurately converting one language to another. The success rate is low, which is why we still use actual translators.

    Solving any one of these three challenges would be impressive and you don’t need expensive to manufacture hardware to prove it. You could easily just release an app on a smart phone to prove your solution.

    What they can probably do is have an okay Microphone in an earpiece that might be able to hear conversations but the likelihood this thing is useful is low. Sounds like a bad Kickstarter project riding the Wearable tech boom.

  2. L Kirk Hagen says:

    “I’d like a cheeseburger, and step on it!”

Post a Reply to L Kirk Hagen

%d bloggers like this: