Reading kana correction model learning device, reading kana correction device, method and program thereof

摘要

PROBLEM TO BE SOLVED: To provide a Japanese kana character correction model learning device and a Japanese kana character correction device, which learn a Japanese kana character correction model as a statistical model for automatically correcting a Japanese kana character error due to notation variability.SOLUTION: A Japanese kana character correction model learning device includes: an N-1 series extraction unit for extracting an N-gram where N-1 characters of hiragana letters appear continuously in one Chinese character in a learning text by using the learning text in which Chinese characters are mixed with kanas, as input; and an N-gram model learning unit for learning a Chinese character-kana N-gram model provided with a probability according to an appearance frequency of the N-gram, and outputting the Chinese character-kana N-gram model as a Japanese kana character correction model to the outside. Also, the Japanese kana character correction model learning device inputs an N-gram of an appropriate Chinese character to the Japanese kana character correction model to find an occurrence probability, and corrects a Japanese kana character of the appropriate Chinese character to a Japanese kana character having an occurrence probability of a predetermined value or more and outputs it.