Learning to Classify Short Text with Topic Model and External Knowledge

摘要: Many methods have been developed to utilize topic analysis models deal with the noises and sparseness of text. However, use a model solely sometimes unable achieve expected high performance, it is very necessary improve current cope characteristic texts specific requirements. In this paper, we focus on two tasks. One make different external corpus identify topics from for better categorization. The other add weight few features in get some those model. We further evaluate performance tasks baseline results. experiments show that our proposed method can higher accuracy text classification. approach find truly representative words which may contribute wide acceptance micro-blog analysis.