Broadening statistical machine translation with comparable corpora and generalized models

As we scale statistical machine translation systems to general domain, we face many challenges. This talk outlines two approaches for building better broad-domain systems. First, progress in data-d