Machine translation has been in development since the early 1950s, but until recently it did not exactly enjoy the best of reputations. However, in recent years there has been a major shift. MT engines now deliver outstandingly high quality thanks to ever-increasing computer performance, enormous quantities of data and progress in artificial intelligence. When combined with professional language service providers, MT therefore has great potential and offers new opportunities for established businesses.
How does machine translation work?
Until a few years ago, “rules-based” and “statistical” engines were used. The former use complex language rules to translate, and the latter are based on translations done by humans. These reference materials are “fed” into machine translation engines so that these engines can learn both terminology and style from the existing translations through the use of algorithms. When an engine translates new texts for the first time, it does so on the basis of the algorithms it has learned. This means that the quality and quantity of the reference materials are key in determining the quality of the machine translation output.
For a number of years “neural” MT engines have been causing quite a stir. For certain language combinations and certain texts they deliver astoundingly high quality.
As MT engines are still not capable of translating with the same quality as human translators, the engine output is often revised and improved by human translators (depending on what the translation is required for). This process is called “post-editing”. The changes made are then added to the engine’s algorithms, and it learns from them. As a result, the more frequently a machine translation engine is used, the higher the quality of the translations becomes and the less post-editing work is required.
MEINRAD.cc has been among the technological pioneers in the translation industry for over 20 years. We will be happy to advise you as to whether MT is useful for your texts.