Translate faster with Machine Translation solutions
As world leaders in machine translation and post edit (humans editing machine translated text) translation we offer the best of both worlds. We can offer custom machine translation solutions or a mix of machines and humans so you get the best quality at the lowest price.
- We can build and train a custom MT engine in most common languages
- Multilingual document discovery – use MT to do a first pass translation then select the parts of the document you would like to have human translated
- Per hour pricing – the better the MT translation the faster the translators go and the lower the cost to you
- Full Post Edit translation service – have your document translated using a mix of machine and humans but get professional human quality output
Want to know more about our range of Machine Translation Solutions? email firstname.lastname@example.org, we promise you’ll get a human and not a machine reply.
How does machine translation work?
We are often asked this question so here is a very quick overview of the MT process.
The machine that produces the translation can only be as good as the data it is trained on, this is the same for all artificial intelligence engines. So the first step is to get some really good existing translations for the language pair and subject you are looking to translate and to use this to train the engine. There are public libraries for the base language pairs and we have specialist data pairs we use on top to get the machines more accurate.
The most common method of machine translation is statistical translation, this uses a powerful statistical engine to predict the correct sequence of translated words based on the data it was trained on. There is also another method that uses vectors to pick the correct sequence and this is often referred to as neural machine translation.
If you are getting a post edit translations then as you translate more content (normally on very large projects) then the translations can be used to refine the engine on the go, making the engine more effective as the project goes on.
It’s also possible to tune the linguist model for the engine based on the subject and language pair which can also improve the engines accuracy.