Exploiting Similarities among Languages for Machine Translation

   Abstract