Machine translation qualifier filter out the raw macine translation There are different kinds of machine translation technologies. Most commonly used classification is made between rule-based machine translation (RBMT) and statistical machine translation (SMT). This division tells how the machine works. However, from the user’s point of view it doesn’t make much difference how the technology makes the translation – as long as the translation is in good quality.

From the quality aspect, machine translation can be classified as raw MT or qualified MT. Qualified machine translation differs remarkably from raw machine translation. Raw or unprocessed machine translation is created by one machine translation service. It is what you get when you use for example Google Translate or Bing Translator. Qualified machine translation is collected by comparing several available automatic translations. Each translation option is being evaluated, estimated and scored automatically. The final translation is being created by selecting the sentences with highest quality scores.

Although machine translation is rarely totally flawless, the difference between the worst and the best available MT is often huge. Everyone who has translated something automatically with Google knows what a really bad translation is. It might be even impossible to get the main idea out of that kind of translation. On the other hand, at its best, MT sounds like it is made by a person who has studied the language as a second language. In practice this often means only minor style or grammar errors. This kind of good MT is a good starting point for post-editing.

MT post-editing is the same as editing automatic translation afterwards. Post-editing raw machine translation can go smoothly but it also can be the quite opposite as well. It’s sometimes easier and faster to build a beautiful house from the beginning than it is to turn a bad building into a dream house. Qualified machine translation is more predictable and stable source for post-editing than raw MT. Basically, qualified MT gives either good quality translation or nothing. Translators don’t have to waste their time on post-editing rubbish.

Raw MT can be useful in personal use when the goal is to understand the key message in a foreign text. For professional use, qualified MT eases and speeds up the translators’ work by creating qualified pre-translations to work with. MT is not supposed to be the final outcome of a translation project. It’s only the first step, and qualified MT is the shortcut to move faster to the top – without leaving any steps in between.


Qualified machine translationDid You Know? Professional translators can improve productivity by 11-50% when they use Qualified Machine Translation.