Friday, 18 January 2013

Lost in translation – voice recognition software still has a long way to go…

An article published by BT last week talks about a new translation app, which claims to allow the user to speak in 23 languages then interprets this across 63 languages (and all for 69p) - this has really sparked some debate here at i-interpret4u.  It all sounds great on paper until you dig a bit deeper into the many issues these kinds of translation tools tend to face.

Yes, Brits are notoriously bad at learning other languages, and yes, we could all do with a bit of a helping hand when it comes to communicating overseas, but in most cases we also need the security and peace of mind that if we simply HAVE to communicate in another language that we absolutely get it right and we absolutely understand what is being said to us.  After all, this could be a make or break emergency situation or something really serious which could put us in danger should we get it wrong.

But BT’s article hits the nail on the head (in the final paragraph) where it admits that this software cannot yet offer perfect interpretations.  What does this mean? Quite simply, it means that the capability of voice recognition technology still has a long way to go before it is truly usable and 100% reliable; it’s a nice thought and a growing technology for sure, but currently it’s far from perfect. 

The problem voice recognition technology faces is that it will never be able to cope with different accents/dialects, if you consider Scots, Welsh, Brummies and Geordie’s alone, you have a whole host of varying accents; but add to that the fact that there are around 45 different Arabic variations and several different Spanish and Portuguese dialects, not to mention speech impediments – then you are looking at technological advancements that are way beyond our years.

To add to the complexity of this, we also have homonyms (words with multiple meanings) to consider: 

·         bow – a long wooden stick with horse hair that is used to play certain string instruments     such as the violin
·         bow – to bend forward at the waist in respect (e.g. "bow down")
·         bow – the front of the ship (e.g. "bow and stern")
·         bow – the weapon which shoots arrows (e.g. "bow and arrow")
·         bow – a kind of tied ribbon (e.g. bow on a present, a bowtie)
·         bow – to bend outward at the sides (e.g. a "bow-legged" cowboy)
·         bough – a branch on a tree. (e.g. "when the bough breaks...")
·         bō – a long staff, usually made of tapered hard wood or bamboo
·         beau – a male paramour

So, after all that we might end up with something completely off track like:

‘The conductor took a tied ribbon after the performance’

Background noise is also another consideration.  Many of these kinds of apps stipulate that for best results you must make sure there are no extraneous noises around you and that your internet connect is stable during voice translations, two important elements that are very much out of your control and certainly aren’t much use if you happen to be in a restaurant, at a hospital, in a taxi or at the airport (many of the places where you might have the need for such a service).

Using apps like this might seem like a cheap, quick hit but if you can’t realistically use it, you’re no better off that you were with your local phrase book.  Voice recognition software is no match for a ‘live’ interpreter where you are guaranteed of the correct interpretation regardless of noise, accent or wi-fi connection.

No comments:

Post a Comment