Paper: Can The Internet Help Improve Machine Translation?

ACL ID N06-3003
Title Can The Internet Help Improve Machine Translation?
Venue Human Language Technologies
Session Doctoral Consortium
Year 2006

This paper summarizes a largely automated method that uses online post-editing feed- back to automatically improve translation rules. As a starting point, bilingual speak- ers’ local fixes are collected through an online Translation Correction Tool. Next, the Rule Refinement Module attacks the problem at its core and uses the local fixes to detect incorrect rules that need to be re- fined. Once the grammar and lexicon have been refined, the Machine Translation sys- tem not only produces the correct transla- tion as fixed by the bilingual speaker, but is also able to generalize and correctly trans- lates similar sentences. Thus, this work constitutes a novel approach to improving translation quality. Enhanced by the reach- ing power of the Internet, our approach be- comes even more releva...