Source-context Features for English-to-Czech Machine Translation
Kamil Kos, Mark Dras · 2009
Statistical machine translation (MT) has become one of the major develop-ment streams in MT in recent years. Models taking advantage of source con-text information have shown that they can improve translation quality. In this project, we investigate the impact of source-context features on the quality of English-to-Czech machine translation. The context we consider are surrounding words and part-of-speech tags, local syntactic structure and other linguistically motivated features that can be extracted from the source language sentence. We implement an extension to the open source MT system Moses, which is used as baseline for our experiments.