摘要 |
PROBLEM TO BE SOLVED: To learn a model for applying a label accurately to an extraction object character.SOLUTION: A dictionary collation part 36 applies a label to a character of second learning data to which the label is not applied. A transition learning identity creation part 38 extracts identities from first learning data and second learning data, and creates an identity vector formed of a target domain identity, other domain identity, and common domain identity about the first learning data and second learning data. A model learning part 40 learns a parameter to each of the target domain identity, other domain identity, and common domain identity, and extracts the parameter to the common domain identity, for learning the model for applying the label accurately to the extraction object character. |