Automatic Text Summarization by Providing Coverage, Non-Redundancy, and Novelty Using Sentence Graph

Background

In a graph based summarization model, a node can be scored using information from the global graph. First, Mani and Bloedorn (1997) proposed graph representation of text. They described a new mechanism for summarizing the similarities and differences between a pair of related documents. Another work used the knowledge of text structure for producing summaries by automatic passage extraction (Salton, Singhal, Mitra, & Buckley, 1997). The earlier iterative graph algorithms are TextRank (Mihalcea & Tarau, 2004) and LexRank (Erkan & Radev, 2004). They can be applied to the summarization of a single or multiple documents in any language (Mihalcea & Tarau, 2005). Even though they are the best graph ranking algorithms, they have high time complexity. Some recent graph based ranking research works are Calvo et al. (2018), Feiyue and Xinchen (2018), and Tixier et al. (2017).