A new preprocessing phase for LSA-based Turkish text summarization

Abstract

Text Summarization is a process of identifying the most salient information in a document or a set of related documents. This paper presents the performance analysis of a Turkish text summarization system that applies two Latent Semantic Analysis based algorithms with different preprocessing phases. The preprocessing method called "Consecutive Words Detection" is a new method that uses Turkish Wikipedia links to represent related consecutive words as a single term and improves the performance of text summarization in Turkish. © 2012 Springer-Verlag GmbH.