| 摘要 |
PROBLEM TO BE SOLVED: To provide a k nearest class deciding device which solves the problem of idiom and the problem of the overlap of classes and can be applied to a natural language processing application. SOLUTION: The distance between a test case 14 and the training case stored in a storage means 10 is calculated by a distance calculation means 11. K train cases having the smallest distance based on distance are selected by a k nearest case selection means 12. When the class of the nearest train case having the smallest distance in these selected k nearest cases are entirely different from the remaining k-1 classes, this class is regarded as an idiom, for instance. The class of the nearest train case is defined as a judgement result class 15 and the class is outputted from a class judgement means 13. |
