A multifactorial approach to Korean -(u)m and -ki

This paper takes a corpus-based approach and examines the linguistic properties of two Korean nominalizers -(u)m and -ki. From the Sejong Treebank corpus, all the sentences with -(u)m and -ki are extracted. Twenty linguistic factors are manually encoded into the extracted sentences. Then, all the encoded data are statistically analyzed with (binary) logistic regression. Although we take a monofactorial analysis, we obtain a good statistical model whose C value is 0.956. Through the analysis, the followings are observed: (i) -(u)m and -ki are used with the ratio of 1:9 in Korean, (ii) among twenty linguistic factors, only ten factors are statistically significant, and (iii) not only the verbs which take -(u)m and -ki as a complement but also the verbs which merge with these two nominalizers also play important roles in the determination of nominalizers.