Glr Parser with Conditional Action Model Using Surface Phrasal Types for Korean

Since the first approach [Wright and Wrigley 1991]of combining a probabilistic method into the GLR technique was published, Some probabilistic GLR parsers also have been implemented in which probabilities are assigned to actions of LR parsing tables by using lookaheads or LR states as simple context information of [Briscoe and Carroll 1993], [Kentaro et al. 1998], and [Ruland, 2000] which does not use the stack information of the GLR parser effectively, because of highly complex internal GLR stack. As a result, they have used relatively limited contextual information for disambiguation. [Kwak et al., 2001] have proposed a conditional action model that uses the partially constructed parse represented by the graph-structured stack as the additional context. However, this method inappropriately defined sub-tree structure. Our proposed model uses Surface Phrasal Types representing the structural characteristics of the sub-trees for its additional contextual information.