Leader: “My universal translator doesn’t seem to be working…”

A look at any one of several well-known sci-fi classics on TV and film tends to show up speech as an area where we’re clearly about to make some big strides.

We regularly see people speaking to computers and vice versa, as well as humans speaking to other species via something called a ‘universal translator’. (OK, so we mean Star Trek - maybe only in the original series did they pretend everyone had done a crash course in English.)

Now while some technology first imagined in futuristic books and movies is pretty much upon us now – think of some areas of wireless communication or biotech – speech is an area that continues to disappoint.

Two announcements today highlight this fact. BT has announced it is to use a “synthesised lady” (perhaps better described as a synthesised lady’s voice) to deliver text messages to users as the spoken word.

So far, so good. This is eminently doable, allowing for the odd bit of weirdness where someone types nonsense – though it should be noted there is a claim smilie emoticons will be cleverly translated, which is interesting.

We can only assume a female voice has been used because it just sounds better. Either that or it will pander to a legion of businessmen who wish they had a PA. Services such as Orange’s Wildfire and the speaking clock have taken the same cue in the past.

Then we also heard about a service being trialled by those clever people at NEC in Japan. One of that country’s long-time computing and electronics giants, NEC is using a room at Toyko’s Narita Airport (where else!). Demonstrations of the Travel Interpreter – part of an 'e-airport’ project - are being held in a special booth with the aim of commercialisation of the service by the end of this year.

Our bet is that won’t happen. While we have no doubt about the quality of software algorithms NEC and others use for speech this is just a damned hard thing to pull off.

To start with, Travel Interpreter will go from Japanese to text to a translation engine then back to speech in English. Or vice versa. All in less than a second, it promises.

As if that isn’t hard enough to do, imagine a fairly average conversation. Remember, this is about interpreting speech, not translation of the written word. Whereas the verb comes near the beginning of sentences in English, in Japanese it’s at the end. Something like ‘There are strange businessmen in this terminal’ in English will read in Japanese as ‘This terminal in strange businessmen there are’. Imagine the process for a really long sentence where what the technology is waiting on to start is at the end of a sentence.

But even if this trial had been limited to languages with similar sentence structure, say English with French or Japanese with Korean, it wouldn’t be more than an interesting experiment for now.

The days of seamlessly integrating speech technology in our everyday lives, or even our business lives, aren’t exactly hurtling closer. But this is a start captains Kirk, Picard, Janeway et al would approve of.