Towards More Effective Text Summarization Based on Textual Association Networks

This paper proposes new text summarization approaches based on textual unit association networks. Textual units refer to words, phrases, sentences, or paragraphs. Intuitively, textual units containing much co-occurrence information are semantically more salient in a document. We construct two kinds of textual association networks, namely, word-based association network and sentence-based association network. For the former, we propose anew approach to computing the word weights and sentence weights.For the latter, we develop a new scheme to score each sentence based on its co-occurrence information. Extensive experiments on benchmark data show that our proposed approaches can achieve better summarization performance than the existing methods. Our approaches are unsupervised, independent of languages, and efficient for different text genre.