Technology Corner

Barriers to Communication

Although the Internet is bringing down the barriers that divide the world, there is still one thing that stands firmly in the way. Language. Despite the global technological revolution, most of us are still unable to talk to others in their native tongue if it is different from our own. Translators are employed around the world, and their services are essential for everything from world peace treaties to technical manuals for toasters. It seems to me to be a logical step to use computers to aid this task.

Computers can now perform extremely complicated spell checking and grammatical operations on even the most detailed of documents. However, we still cannot translate them into another language with one hundred percent accuracy. While at present there are document translators in operation they have extreme problems with certain types of sentences. I will give you an example. "Hydraulic rams are used to support tipping trailers." The computer would have a great deal of trouble interpreting this accurately. In fact one early translation program defined Hydraulic Ram as "Water - sheep". Perhaps you can now see my point. So, has this little problem actually deterred the developers ?

Not at all.

Most of the large developers have had a go at the problem. In the mid 50's, IBM came up with a translator that could translate Russian sentences into English. Needless to say its success was limited. But nevertheless it worked to an extent and provided some small basis upon which to build future systems. It was not long before companies began to obtain translating systems and then use human translators to touch up the document when the computer had finished. It was argued, mainly by the human translators, that this process took almost as long as it had originally taken them before the introduction of the system.

Extensive research and work was still being lavished on the Computerised Translators and by the late eighties the accuracy of the system was well into the ninety percent range. You can buy pocket translators that will ask you for a word in English and then translate into the nearest French equivalent.

The question is how far have the developers managed to get. Wel,l the fact is that computers are getting quite fast at the translation process. While it would take a human almost a day to convert a ten thousand word document in a foreign language, it can take a computer a mere forty minutes to provide a basic translation. This, as you can imagine, is a necessity for organizations such as the EEC who need to convert documents into many different languages with the minimum amount of fuss. The accuracy of these systems is still in the high eighties and the translators often contain a vocabulary more than 36,000 words. It is only when it comes to the context of some words that the systems do still experience some problem. I think that it is unrealistic to think that this will ever be fully rectified. I think that it is more likely that we will modify our speech to allow better translations. We already do this unconsciously when we speak to someone who can only speak a little of our own language.

Of course, the next step in this process is to be able to translate the spoken word. There are systems around already that can recognize speech, some of which I have mentioned in previous articles. However, it is an extremely large step for a computer, which is already struggling to recognize your speech, to then go through the motions of translating. Still, even this challenge has not been ignored. Neural Networks are being used to aid the computer with the more lateral task of recognizing things like phraseology and speech.

British Telecom PLC are working on a system that can translate speech. At the moment it is fairly slow and not incredibly accurate. However a similar system is already in use in some large Japanese Hotels to take reservations and bookings.

In the short term I think that we will increasingly come across voice response systems operating within large corporations. Their task will be to route data and communications traffic to its correct destination. In the long term I have no doubt however that this technology will have improved to a point where it could be used to translate live dialogue with a good degree of accuracy.

I am sure that I do not need to elaborate on what the success of such a system could do for the people of this planet. But one thing is certain. It would bring the communication barriers tumbling down, leaving us without an excuse for silence.

Front Page