한국어의 두 목적어 어순에 관한 코퍼스 연구

Abstract

This study aims to identify the canonical word-orders when Korean verbs take two objects: a clause and an NP. To pursue this goal, with over 7,000 tokens of 27 Korean utterance verbs which reportedly possess a ‘hearer-theme’ construction (Cho 2009; Lee 2016), it observes how many objects (zero, one, or two objects) are realized in a token sentence, analyzes in what type of syntactic categories the objects appear, and finally examines which word-order is preferred in case of the two objects in a token. Besides, it attempts to classify the 27 utterance verbs in terms of their object realization, i.e., the number and the syntactic types of objects in the tokens, using K-means clustering algorithm and Heirarchical Clustering. The findings are (i) the most of the data fell into zero or one object cases while only 5% of the data occurred with two objects, (ii) the canonical word-order differed depending on the syntactic type of an NP object so that an Accusative NP mostly followed a clausal object while a Dative NP preferably preceded an object clause, and (iii) the verbs given were best categorized into three groups when they took only one complement: hearer-dominant group, theme-dominant group, and hearer/theme group.