MODEL LEARNING METHOD AND METHOD AND DEVICE FOR DESCRIPTIVE TEXT EVALUATION

摘要

PROBLEM TO BE SOLVED: To evaluate a descriptive text for a document with good accuracy.SOLUTION: An identity extraction unit 30 extracts from each descriptive text for learning: an identity regarding existence of an image link; an identity regarding existence of a key word; an identity regarding structure, an identity regarding dispersal expression expressing a meaning of a word; an identity regarding whether or not a keyword included in the descriptive text for learning is included in the descriptive document for learning; identity regarding similarity between dispersal expression of a word included in the descriptive text for learning and the dispersal expression for the word included in the document for learning; identity based on contribution time for the descriptive text for learning and the contribution time for the document for learning; an identity regarding a user attribute included in the contribution information for learning; and a plurality of identities including the identity regarding the inclusion of the key word in the descriptive text for learning in the meta data for learning, and a model learning unit 32 learns an evaluation model based the plurality of identities for each descriptive text for learning and evaluation values for learning regarding each descriptive text for learning.SELECTED DRAWING: Figure 1