With the increasing growth of digital documents, text categorization has become an useful technique for organizing text data. In text categorization, term-weighting methods which assign an appropriate weight to each term are usually utilized to improve the classification performance. In this work, we propose two entropy-based weighting schemes. Using the category-distribution information of terms, these schemes to weight the terms based on their certainties of distribution in categories. The experimental results demonstrate that the proposed schemes outperform existing methods in improving text classifier. where is the end.