New Parsing Method Using Global Association Table

This paper presents a new parsing method using statistical information extracted from corpus, especially for Korean. The structural ambiguities are occurred in deciding the dependency relation between words in Korean. While guring out the correct dependency, the lexical associations play an important role in resolving the ambiguities. Our parser uses statistical cooccurrence data to compute the lexical associations. In addition, it can be shown that sentences are parsed deterministically by the global management of the association. In this paper, the global association table(GAT) is deened and the association between words is recorded in the GAT. The system is the hybrid semi-deterministic parser and is controlled not by the condition-action rule, but by the association value between phrases. Whenever the expectation of the parser fails, it chooses the alternatives using a chart to remove the backtracking.