Poster ID : M5

Semi-supervised Learning with Weakly-Related Unlabeled Data: Towards Better Text Categorization
Liu Yang, Rong Jin, Rahul Sukthankar
Reuters
weakly related

WebKB

The cluster assumption may not be valid !

· What if unlabeled data weakly related to target classes · Estimate the optimal word-correlation matrix · Text categorization on a small training pool of Reuters