Low tech print to high tech AI

From printing blocks to AI 

Using computers to translate between languages is nothing new. The first public demonstration of a machine translation system took place in 1954. The IBM technology was, understandably, very basic, translating between just 250 words of Russian and English. However, it caused quite a stir, raising expectations of what machine translation could achieve in future. Sixty-five years on, are computers now ready to replace human translators? We’re all familiar with free website translation services, such as Google Translate. It was in very recent memory that these tools offered little more than gibberish translations. But thanks to major technological advances, they’ve become pretty usable. Artificial intelligence (AI), in particular, has revolutionised machine translation. Sophisticated modern machine translation systems use neural networks – computer systems inspired by the brain. These move away from ‘rules-based’ systems which work by breaking down the rules of one language and then reconstructing them in another. Neural networking, on the other hand, allows machines to constantly ‘learn’. By repeatedly analysing huge swathes of text it can spot patterns and ‘predict’ which words are likely to go together.  This leads to more accurate and natural sounding translations.

Machine translation services which use neural networking include Google Translate, Amazon Translate and Microsoft Translator. While their work still needs a check-over by humans, such tools have some advantages over the real deal. For instance, they may be useful if the text needs to be FULLY confidential. They are also faster and often free. Furthermore, you can carry them around in your pocket courtesy of your smartphone – handy when you’ve not got a human translator to hand. However, translation is a sophisticated art and should be able to deal with the ambiguities, nuances and emotion of human communication. Humans win here.

Humans are much better at understanding how context affects meaning.  They can analyse all elements of a text and know how they influence each other. Machine translation will currently only translate one sentence at a time and will make mistakes if the meaning of a sentence depends upon another. They also struggle with rare words and long sentences. Translation software is continually improving, and there is plenty of scope for machine and human translators to work together. However, for fully professionally translation services, especially in client and market facing situations, it will be some time before computers can truly replace human translators.

At Bath Translations our friendly team understand our clients need for a professional and accurate service. Should you have a translation, interpreting, voice-over or transcription enquiry. contact us for advice by email or calling 01985 250074 for enquiries. We are happy to help.