Compound Noun Analysis for Process of Korean Unregistered Word
Guanghe Jin, Zhanguo Li, Dapeng Qu, Xingwei Wang · 2012
In this paper, a new method of compound noun analysis is proposed. It uses decomposition model and unregistered words recognition. The latter contains loanword nouns, name nouns and place name. Loanword noun is recognized based on it's formed by syllables which have low use frequency in Korean syllable. Name noun is recognized based on the characteristics of Korean name and the repeatability of the certain amount of syllables. Place name is recognized by consulting place name dictionary based on the analysis of model that the place names appear. The experimental results for compound noun analysis have an average accuracy of 98.2%. The accuracy of loanword nouns recognition and name nouns recognition is 92% and 95% separately.