Our post-editing services have been good enough to be approves as a supplier by the European Union when they used rule-based machine translation services.
There are three levels of service in post-editing
- Raw MT output
- Light post-editing
- Heavy post-editing
Raw post-editing is the very output from the translation software without any human intervention. The software may produce good enough sentences that are understandable, and even perfect to understand for the intended audience. Some may be easy to understand although they are not grammatically correct. Others will be somewhat harder to understand although the meaning can be discerned from the words in the sentence. A small proportion of sentences (typically below 10%) cannot be understood.
A typical use case from one of our clients is our development for Sybase, reported by that corporation at Localization World 2011
Light post-editing is a service for medium or low-value content that is intended for internal use, not for publishing. A translator quickly reads the text to correct the sentences where machine translation has failed in order to offer human translation. The translator does not correct style or sentences that can be understood by a human with some effort. Overall, the document looks like machine-translated but any sentences that cannot be understood are improved.
Heavy post-editing is a service for medium-high or high value content that is intended for final publication. Normally, heavy post-editing is used by professional translators who scrutinise the text comparing sentence to sentence between the source language and the target language. The translators aim to offer a final version after post-editing which reads and sounds as if originally translated by a human translator. The final document is good for publishing.
Heavy post-editing is still somewhat cheaper than human translation services as part of the job has been previously done by a translation software. A second pair of eyes may be required, depending on the level of perfection or quality needed by the client (final proofreading). The translator corrects typical grammatical errors coming from the software, style or sentences that otherwise would not be understood by humans without great effort. Some sentences may be re-written from scratch.
Some translators use MT this way in order to have a first draft of the sentence or paragraph they are about to translate, and apply their own terminology databases and experience in order to speed up the translation process.
We can provide samples of all three cases (raw machine translation, light post-editing, heavy post-editing) upon request as well as some screenshots of PangeaMT at work. You can see a video of the platform’s capabilities visiting our video page.