Under the right situations, machine translation, sometimes referred to as MT, is a powerful tool. Delivering low-quality conversion of text within seconds is an impressive feat compared to humans taking hours or days to work through a large document. In this article I’m going to share with you a guide on how machine translations work so that you can envisage how best they can work for you.
Previous versions of machine translations depended upon a ‘dictionary’ and the use of grammatical rules for the particular language being used. Times have changed and so has the approach of MT. They now operate by using a range of example translations. This modern approach means that they are able to produce better results by using statistics to convert one language to another. For example if there are particular words in a sentence, then what chance is there that these same words will appear in previous translations and in a particular structure.
Having such a large body of samples to work from, this machine translation system can translate by putting words together that are statistically more likely to be in a certain structure.
Many people are surprised that this is how machine translation works. They are looking for grammatical analysis and rule based systems. However, this outdated approach makes it more likely for there to be errors. Grammar does not always follow the same rules, especially in the English language – as we already know.
So what do people want to use machine translations for? Quite often they want to look up just a single word or a small piece of text if they are using it online. If they approach one of the machine translation service providers, then typically they will have a large document that has technical words in it. But in most cases they are looking to use MT for giving a gist of what is meant by a particular text.