Beyond Bitext: Five Open Problems in Machine Translation
Adam Lopez, Matt Post · 2013
In twenty years, the machine translation (MT) research community has learned a great deal about problems that can be solved with bi-text. Yet for many potential MT uses, there is little if any available bitext. In the next twenty years, these uses will become increas-ingly important, and the research community must marshal its resources to solve the new problems that they present. Specifically, we must assemble large numbers of small bitexts for testing systems, rather than small numbers of large bitexts for training them. Small bi-texts won’t solve the new problems alone, but they will help the research community identify