“To boldly go where no man has gone before.” Ouch! That split infinitive makes me cringe every time I hear it. And as a fanatic of the Star Trek franchise, I must have heard it over a thousand times.
Couldn’t they have proofread the script before sending it in for recording? Did nobody notice?
The final frontier for developers of translation technologies seems to be a world where speech is instantly translated and spoken back in any desired language. In the fictional 23rd century of the Star Trek Original Series, everyone, even the little green men, speak flawless (well, almost flawless) American English. And the only explanation they offer for this monolingual universe is the computer’s universal translator, a machine capable of atomizing any language into the smallest meaningful bits, beaming them through its circuits, and rearranging them into English.
Today, the technology necessary for someone to build such a machine is readily available. Machine translation has been around for decades, and so have speech recognition and speech synthesis technology. Moreover, since the advent of smartphones, these technologies are quite literally in everyone’s pocket.
So what will it take to achieve this universal translator?
The challenges ahead for developers of translation and interpretation technologies is best illustrated by a humorous mishap which occurred in 2012, after the worldwide release of Siri, a voice activated factotum that lives inside the iPhone. The problem was that the speech recognition software could not understand the Scottish accent. Hilarious videos of Scotsmen struggling to give Siri simple commands such as, “Please find a pub in Aberdeen” went viral. To me, these incidents were reminiscent of the scene in Star Trek IV, The Voyage Home, where Scottie travels back in time to 1986 and attempts to communicate with a Mac Plus, pleading “Hello, Computer! Hello Computer!” to no avail. And here, we are not talking about little green men, or even inhabitants of remote corners of the globe, but speakers of English, arguably the most commonly spoken language on Earth.
The developers’ continuing mission, then, is to reach beyond regional boundaries and adapt machines to recognize the speech patterns that make up the differences in accents. A database containing the tendencies and particularities of, say, German, may be capable of understanding Germans speaking English, Spanish, Lao, or even Burmese. Inflection, tone, and common mistakes are all perfectly measurable and quantifiable. The next development in these technologies may well be a small step/giant leap moment.