Models for Inuktitut-English Word Alignment

摘要

This paper presents a set of techniques for bitext word alignment,optimized for a language pair with the characteristics ofInuktitut-English. The resulting systems exploit cross-lingualaffinities at the sublexical level of syllables and substrings, aswell as regular patterns of transliteration and the tendency towardsmonotonicity of alignment. Our most successful systemswere based on classifier combination, and we found differentcombination methods performed best under the target evaluationmetrics of F-measure and alignment error rate.