Korean Compound Noun Decomposition Using Syllabic Information Only

The compound nouns are freely composed in Korean, since it is possible to concatenate independent nouns without a postposition. Therefore, the systems that handle compound nouns such as machine translation and information retrieval have to decompose them into single nouns for the further correct analysis of texts. This paper proposes the GECORAM (GEneralized COmbination of Rule-based learning And Memory-based learning) algorithm for Korean compound noun decomposition using only syllabic information. The merit of rule-based learning algorithms is high comprehensibility, but they shows low performance in many application tasks. To tackle this problem, GECORAM combines the rule-based learning and memory-based learning. According to the experimental results, GECORAM shows higher accuracy than rule-based learning or memory-based learning alone.