Korean Compound Noun Indexing Based on Lexical Association and Conceptual Association

Conventional methods have dealt with compound nouns with a goal to enhance the recall. That is, they've extracted only unit nouns within a full compound noun but didn't take out head modiiers which can improve the precision. This paper presents a new method of the Korean compound noun indexing which can improve both the recall and the precision. Our method extracts head modiiers from a compound noun by using lexical association and conceptual association. To enhance the recall, the unit nouns of composing compound nouns are extracted as indexing terms. To increase the precision, head modiiers within a full compound noun are obtained by the weighted sum of lexical association and conceptual association. In the experiment, we try to evaluate the eeectiveness of using head modiiers to supplement single nouns for indexing. The experiment data are 2,600 documents and 30 queries in KTSET 2.0. Experimental results show that the proposed method yields a retrieval performance better than previous techniques.