... Data point selection The key idea in our experiments is that we can use asimple form of instance weighting, similar to what isoften used for correcting sample selection bias or for domain adaptation, ... remaining data we use. For example, for Bulgar-ian our baseline result using 100% of the source datais 44.5%, and the result obtained using 90% of thesource data is 70.2%. Using held-out data, ... the Association for Computational Linguistics:shortpapers, pages 682–686,Portland, Oregon, June 19-24, 2011.c2011 Association for Computational LinguisticsData point selection for cross-language...