Recently,with the incteasing availability of large corpus,there have been numerous attempts to automatically acquire the linguistic information from the corpus.Syntactic relations are one of such information that can be extracted automatically from a corpus and is very useful.But the case shift and ellipsis phenomena,duplicated usage of the compound word in Korean make the automatic extraction difficult and the problem of sparse data more critical.This paper tries to solve these problems through a corpus normalization made by reflecting grammatical features in Korean. Keywords:Corpus,Information Acquisition,Case shift