A word alignment system based on a translation equivalence extractor
Romanian Academy Center for Artificial Intelligence
The paper describes a word alignment system (TREQ-AL), which uses the lexicon extracted with a translation-equivalents extractor (TREQ) from a training corpus, including the text to be aligned. The improvement methods applied since the previous version of TREQ-AL are one of the paper's focus, as well as the recalled types of the system.
parallel corpus, word-to-word alignment, linguistic methods, translation-equivalents