A Method of Reducing the臼ndidatesof 'Kanji -Kana' Strings Translated from the Non-segmented 'Kana' Strlngs Using Markov Cbain Models of Cbaracters and Words

摘要

There are many ressearches on the method which translates the non-segmented "Kana" sentences into the "kana"sentences. However，the amount of computer memories required for the translating processing explodes in many times， because the number of combination of candidates for ''kanji -kana" words grows in proportion to the increasing of the length of the sentence. The memory explosion can be prevented if a sentence is separated into "bunsetsu". Up to now，an useful method for finding and correcting the provisionalboundaries，of "bunsetsu" using 2nd-order Markov model has been proposed. This paper proposes a method of reducing the 'bunsetsu" candidates of "Kanji-Kana" strings translated from the non-segmented ''kana bunsetsu"， using Markov models of character and word.