Reinforcing Parser Preferences through Tagging

Lexical ambiguity is an important source of inefciency for wide-coverage HPSG parsing. In this paper, we propose a lexical analysis lter which removes unlikely lexical cat- egories. The lter is implemented as a straightforward HMM n-gram POS-tagger, which com- putes the 'a posteriori' probability of each lexical category. A lexical category is removed if a competing lexical category is sufciently more likely. The novel aspect of our approach is the fact that the tagger is trained on the output of the parser itself; therefore there is no need for hand-annotated material. Use of this lter increases the speed of the parser considerably, and in addition gives rise to an improvement in parsing accuracy. R · ESUM · E. L'ambigu¤ · e lexicale est une source importante de l'inefcacit · e de l'analyse syn- ecision.