Extracting discriminative concepts for domain adaptation in text mining