Statistical Dependency Parsing in Korean: From Corpus Generation To Automatic Parsing

Abstract

This paper gives two contributions to depen-dency parsing in Korean. First, we build a Ko-rean dependency Treebank from an existing constituent Treebank. For a morphologically rich language like Korean, dependency pars-ing shows some advantages over constituent parsing. Since there is not much training data available, we automatically generate depen-dency trees by applying head-percolation rules and heuristics to the constituent trees. Second, we show how to extract useful features for dependency parsing from rich morphology in Korean. Once we build the dependency Tree-bank, any statistical parsing approach can be applied. The challenging part is how to ex-tract features from tokens consisting of multi-ple morphemes. We suggest a way of select-ing important morphemes and use only these as features to avoid sparsity. Our parsing ap-proach is evaluated on three different genres using both gold-standard and automatic mor-phological analysis. We also test the impact of fine vs. coarse-grained morphologies on de-pendency parsing. With automatic morpho-logical analysis, we achieve labeled attach-ment scores of 80%+. To the best of our knowledge, this is the first time that Korean dependency parsing has been evaluated on la-beled edges with such a large variety of data. 1