Context-Sensitive Detection and Correction of Homonym Errors in Japanese Texts

摘要

In Japanese and Chinese, the use of kanji characters has produced words with many homonyms. Thus, the correct 'spelling' (kanji) for a word can depend on its context. Computer text entry systems provide a mechanism for users to choose between possible spellings, but mistakes are relatively common. We propose a new context-sensitive algorithm to correct this kind of spelling error. The algorithm detects errors in input text and suggests alternative spellings to the user. Our experimental results have shown maximally about 98% of accuracy rate.