Bitext Alignment J?rg Tiedemann

摘要

Bitext alignment techniques are at the heart of the revolution that swept through the field of machine translation (MT) two decades ago. From the early attempts of pioneers like Kay and Réscheisen (1993), to the recent success of [enter name of your current favorite alignment method or researcher], the growth of translation alignment is so tightly intertwined with that of statistical machine translation that it is actually difficult (and most likely futile) to try to establish which was most instrumental to which. From the beginning, researchers have been attracted by the alignment problem, not only because of its essential link to MT and the numerous other applications that could be derived, but also because bitext alignment seems like a “neat” problem: one around which a researcher can easily wrap his or her head until a clean solution emerges.