Битага Модель Скрытой Марковки

We combine two complementary ideas for learning supertaggers from highly am- biguous lexicons: grammar-informed tag transitions and models minimized via in- teger programming. Each strategy on its own greatly improves …...