Construction of Korean Semantic Annotated Corpus

Abstract

Semantic role labeling is to determine semantic relationships between a predicate and its arguments in a sentence. Although semantic annotated corpus should be necessary to do the labeling, unfortunately, Korean language is lack of the corpus data. In this paper, we construct Korean Proposition Bank (PropBank) corpus and Sejong case frame annotated corpus. Semantic annotation means a task which gives a semantic tag to an argument word in consideration of the dependency relationship between a predicate and the argument. And semantic annotated corpus is used in semantic role labeling based on machine learning methods. We annotate the syntactic annotated corpus built by Korea Electronics and Telecommunication Research Institute. We use frame files for Korean PropBank and Sejong case frame dictionary for Sejong case frame annotated corpus.