Most of Korean morphological analyzers generate more morphological analyses then we need. This is one of the reasons to make difficult for Koreans processing systems such as a part-of-speech tagging system and a syntactic parsing system. To reduce the...
Most of Korean morphological analyzers generate more morphological analyses then we need. This is one of the reasons to make difficult for Koreans processing systems such as a part-of-speech tagging system and a syntactic parsing system. To reduce the morphological analyses, we use two kinds of lexical information as linguistic knowledge; one is called the lexicalized morphological and the other is called the subsumption relation on morphotactics analyses. Our experiments shows that the lexicalized morphotactics and subsumption relation are very useful linguistic information to reduce the morphological ambiguity. When using two information in Korean morphological analysis, the reduction rate of the morphological structures is 68%.