News Summarization Based on Semantic Similarity Measure

Abstract

This paper proposed a new method of news summarization based on semantic similarity measure. It used Latent semantic indexing (LSI) to measure sentence similarity, then it used Singular Value Decomposition (SVD) to reduce the dimension of the word-sentence matrix, it used new clustering algorithm to cluster all the sentences. It ordered all the sentences according to their relevant positions in the original document. Experimental result shows that the proposed method can improve the performance of summary.