Exploiting Category Information and Document Information to Improve Term Weighting for Text Categorization