| 摘要 |
PROBLEM TO BE SOLVED: To learn the correction of an error by determining the probability that words in a target text are correct words according to the appearance of words in a combined training corpus composed of an ordinary training corpus and the target text in combination, and correcting spelling. SOLUTION: In a training stage 12, the target text 16 and ordinary training corpus 18 are combined by a combining means 20 into the combined training corpus 21. In an execution time processing state 14, when the target text 16 is prepared, a correcting means 30 determines a target word first and then determines the probability that the target word is a correct word according to the combined training corpus 21 and features 24. For the correction of spelling by the correcting means 30, the combined training corpus 21 is utilized and when the target word needs to be replaced, a word to be inserted into the text instead of the target work is proposed as the result of the spelling correction based upon the probability of the correcting means 30. |
