Automatic Detection of Machine Translated Text and Translation Quality Estimation

We show that it is possible to automati-cally detect machine translated text at sen-tence level from monolingual corpora, us-ing text classification methods. We show further that the accuracy with which a learned classifier can detect text as ma-chine translated is strongly correlated with the translation quality of the machine translation system that generated it. Fi-nally, we offer a generic machine transla-tion quality estimation technique based on this approach, which does not require ref-erence sentences.