PROFILE ESTIMATION MODEL LEARNING METHOD, PROFILE ESTIMATION METHOD, DEVICE AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To allow learning of a profile estimation model at high speed.SOLUTION: A learning word notation origin creation part 22 creates a word notation origin for each of learning documents which are created by plural learning users. A learning part 23 learns plural learning models on the basis of teacher data and the created word notation origins. A co-occurrence probability calculation part 24 between profiles calculates co-occurrence probability for each combination of plural profile attributes on the basis of the teacher data. A score correction part 25 corrects each calculated co-occurrence probability so as to correspond to a magnitude relation of a value which indicates relationship between the word notation included in the plural learning models and attribute value of the profile attribute, and a range of the value. A model coupling part 27 creates a profile estimation model on the basis of the learned plural learning models, and each corrected co-occurrence probability.