This is a subject that fascinates our industry – the idea of a machine working autonomously to convert a text from language to another. Quick and painless. So I dedicated my MA to research this subject.
Machine translation (MT) results depend on a combination of type of MT system, languages and text domain. The more strict is a domain, and the more alike are the languages, the more accurate – or demanding less post-editing effort – the results will be. Training the local systems and using a corpus with a generous size can also impact the translation.
My dissertation deals especifically with package inserts and several MT systems. The text is available in Portuguese only, but in case you’re interested in any chapter specifically, just let me know and I’ll provide you a translation into American English.
You can download it here.