A Korean Morphological Analyzer using a Pre-analyzed Partial Word-phrase Dictionary
Shin Joon Choul, Cheol-Young Ock · Jeongbo gwahaghoe nonmunji. so'peuteuweeo mich eung'yong · 2012
The Korean morphological analysis based on corpus usually uses the pre-analyzed full word-phrase dictionary(FWD) that is constructed from the corpus. If input words are not found in the FWD, the morphemes of the input words are analyzed using complicated analysis rules: code transformation, decomposition of morphemes, and restoration of original form. Such complicated analysis rules are inefficient in terms of programming, maintenance, and runtime and cause to reduce its accuracy and performance. In order to solve these problems, the method using a pre-analyzed partial word-phrase dictionary(PWD) was researched. This paper proposes new method that constructs the PWD from tagged corpus and analyzes Korean morpheme using the PWD. According to the experiments on Sejong corpus, the recall of morpheme analysis is 99.05%. And the accuracy of POS with homonym tagging based on Hidden-Markov-Model is 96.76%.