JAPANESE READING SYLLABARY ERROR DETECTION DEVICE, METHOD, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide a technique for evaluating naturalness of an input text as Japanese on the basis of both connection between Japanese reading syllabaries and connection between a kanji character and a hiragana character to automatically detect a part in which a Japanese reading syllabary error has highly possibly occurred.SOLUTION: A Japanese reading syllabary error detection device comprises: an unnatural reading syllabary connection detection unit which detects an unnatural reading syllabary connection part in which connection between Japanese reading syllabaries is unnatural in an input text by using a reading N-gram model which statistically learns natural order of reading syllabaries as Japanese; and an unnatural kanji-hiragana connection detection unit which detects an unnatural kanji-hiragana connection part in which connection between a kanji character and a hiragana character is unnatural in the input text by using a single kanji dictionary with connection and reading frequencies composed by calculating a frequency, at which a hiragana character sequentially appears next to a kanji character in a learning text to which correct reading syllabaries are attached, for each combination of the kanji character, the reading syllabary of the kanji character and the hiragana character appearing next to the kanji character.