Paper: Word Confidence Estimation for SMT N-best List Re-ranking

ACL ID W14-0301
Title Word Confidence Estimation for SMT N-best List Re-ranking
Venue EACL Workshop on Humans and Computer-assisted Translation
Year 2014

This paper proposes to use Word Confi- dence Estimation (WCE) information to improve MT outputs via N-best list re- ranking. From the confidence label as- signed for each word in the MT hypoth- esis, we add six scores to the baseline log- linear model in order to re-rank the N-best list. Firstly, the correlation between the WCE-based sentence-level scores and the conventional evaluation scores (BLEU, TER, TERp-A) is investigated. Then, the N-best list re-ranking is evaluated over dif- ferent WCE system performance levels: from our real and efficient WCE system (ranked 1st during last WMT 2013 Quality Estimation Task) to an oracle WCE (which simulates an interactive scenario where a user simply validates words of a MT hy- pothesis and the new output will be auto- matically re-generated). Th...