A New Approach for Multi-document Summarization Based on Latent Semantic Analysis

Abstract

Multi-document summary plays an increasingly important role with the exponential document growth on the web. Among many traditional multi-document summarization techniques, the latent semantic analysis (LSA) is a unique duo to its using latent semantic information instead of original feature, which results in a better performance. However, since those approaches based on LSA evaluate and select sentence individually, none of them is able to remove the redundant sentences. In this paper, we propose a new method to evaluate a sentence subset based on its capacity to reproduce term projections on right singular vectors. Finally, the experiments on DUC2002 and DUC2004 datasets validate the effectiveness of our proposed methods.