HPSG Parsing with a Supertagger

This chapter describes probabilistic HPSG models with a supertagger for fast and accurate parsing. A supertagger is a probabilistic model for selecting lexical entries with word and POS n-gram features. Recently, supertagging has become well known to drastically improve parsing accuracy and speed in CCG, HPSG and CDG parsing.We propose three models for probabilistic HPSG parsing with an HPSG supertagger. In the first model, the probabilities of parse trees are defined with only the HPSG supertagger. This model is very simple, and experiments revealed that the implemented parser ran around three times faster, and had comparable accuracy with the previous probabilistic HPSG parser. The second model is defined as the product of the probabilities of the HPSG supertagger and the previous probabilistic HPSG parser. The second model is not only significantly faster but also significantly more accurate than the previous model. The last model is a log-linear model in which the supertagger is introduced as its reference distribution. This model properly incorporates the supertagging probabilities into parse tree’s probabilistic model, and achieved the best performance among the proposed models.