A high quality MT system has good domain knowledge. This basically means that the MT system knows something about the content it’s translating. For example, if you’re translating medical content, then you should use an MT system that has been ‘trained’ to work with medical content.
This is similar to the way you’ve always worked with translation memories. For example, you’d only use a medical translation memory on medical texts; it improves the perfect and fuzzy matches and ensures a higher quality translation for your clients.
Machine Translation systems are no different. The really amazing thing is that you actually use your translation memories to ‘train’ your MT system. So you can immediately improve the quality and consistency of your KantanMT engine.
The training of MT systems with domain knowledge is a key factor in improving the quality and consistency of MT output.
At Kantan we’ve made it very easy to ‘train’ your MT engine. You simply upload your bi-lingual translation memories into your private and secure data storage area and it will automatically be used to train your MT engine. This improves the quality of the output, its consistency and ensures that you do less post-editing on your client’s content.
So now you see that using MT is similar to using TM, except of course, at Kantan, we translate your new content as well as your old!