A new approach for summarizing documents using generic impressions expressions

With increasing of the availability information on the Web, document summarization technologies become very important research. There are many techniques using latent semantic analysis (LSA) and non-negative matrix factorization (NMF), but there are no researches on matrix reduction and computation speeding for NMF method. Therefore, this paper utilizes the generic knowledge of expressions from newspapers to extract important sentences as the summary without stemming processes and no filtering of stop words. Form experimental results, it turns out that the proposed method reduce the matrix size reduction and speed up the computation of the NMF method without changing on the original relevancy.