This thesis presents a new computational model for morphological analysis of Korean, recovering all the feasible sequences of morphemes of a given input word. We call the model the Prediction-based Morphological Analysis (PMA) model. The PMA model deals with some issues in morphological analysis, mainly concerned with morphological decomposition of inflected/ambiguous multimorphemic words: that is, the systematic and predictive treatment of spelling changes which occur between morphemes for inflected words, and the efficient processing of morphological decomposition for ambiguous words.
The PMA model works with little computational redundancy in processing the multimorphemic words. The computational efficiency is made possible by three types of new techniques: first, predictive (selective) rule application which restricts to the spelling rules suitable for the input word; second, a morphological chart parsing which enables the analyzer to avoid recomputing analyzed substrings in case the input word is morphologically ambiguous; third, a method for interpreting and compiling spelling rules. The PMA model has been experimented with the words selected from the corpus of Korean elementary textbooks. Experimental results show that the model guarantees fast and reliable processing.