Automatic Translation Post Editing
Post editing is a translation technique in which the translator receives computer-generated suggestions to improve quality and productivity.
For the payment of an initial license fee or for translation system training, guarantees a saving on regular translation rates of 5% to 20%.
Since 2002 we have translated more than 5 million words using post editing for clients such as the European Commission, IBM and Google, as well as for numerous small enterprises.
When Is It Used?
Anyone who translates over 500,000 words per language per year and has built up translation memories containing over 1,000,000 words will gain a return on investment in the first year (or sooner for higher volumes).
Training costs range from 1,400 to 7,000 USD depending on how much data needs to be processed. Rate discounts stand at 0.014 USD per word.
We use automatic translation both to offer quality translations and to provide "gisting" services which, at a significantly lower cost than traditional translation, is used to correct only the most common errors.
OK, I Like It. What Do I Do?
Case 1: You have a memory containing more than 1 million words
- Go to http://mymemory.translated.net.
- Log-in, or create a new account if you don't have one.
- Upload your memory in TMX format to obtain your initial training for free.
- Contact us at info@translated.net with your username and with a text to be translated of up to 1,000 words.
- Within 2 weeks, we will send you:
- the document, automatically translated using statistical machine translation
- the version corrected by the translator (post editing), which you may use to evaluate the quality of the end product
- an estimate for the training of other languages and an estimate for post editing.
Case 2: You do not have a memory containing more than 1 million words
The same process, but starting from point 4. We will initially use a generic system as a starting point. Zero startup costs and a small immediate saving.
Our Post Editing Experience
- Translated, in partnership with Systran, is awarded a project by the European Commission testing new models for integrating automatic translation into the human translation process.
- Translated begins developing Semantix, an NLP toolkit containing a statistical translation engine based on the theory of Brown and Della Pietra, supplemented by a set theory-based learning model.
- Translated begins to offer SaaS automatic translation services through an API. It develops its first translation engine based on Giza++ and Pharaoz.
- Translated initiates the MyMemory project: organizing worldwide linguistic information to make it available free of charge in the form of translation memories and automatic translations.
- MyMemory get deployed internally.
- MyMemory reaches the 300 million-word mark, all of which are made available for free online.
- MyMemory exceeds 5 billion words. The project is presented by Marco Trombetti in a keynote presentation at the Machine Translation Summit in Ottawa, Canada. IBM assigns Translated the first of its post editing projects with its own statistical technology. The European Commission does the same. Google develops the Google Translator Toolkit, a web-based cat tool for post editing, and assigns Translated its first internal post editing project: localizing the Toolkit, using the Toolkit itself.
- MyMemory reaches 50 billion words and is directly integrated into all major CAT tools and translation management systems. Our mission: terms should actually be sentences or similar.