Article Preview
Top1. Introduction
In recent years, web has been providing enormous amount of information in every field. Managing of overloaded data is a crucial task. Moreover, the effect of expanding the information in web and advanced libraries can lead to typical issues. One answer to this problem is to shorten the data in form of a concise document through summarization task. The main objective of this task is to reduce the original text without losing main contents (Aliguliyev, 2009). The concise document, called summary gives quick reference to create interest, also helps in decision making and act as time saver for readers. The task of summarization depends on the way it is generated summary such as extractive and abstractive summary. Important part such sentences, paragraphs, etc., are extracted out of the document is called extractive summarization whereas abstractive summarization requires linguistic analysis to generate summary (Binwahlan, Salim, & Suanmali, 2009; Ježek, 2008; Lloret, 2012; Mendoza et al., 2014; Oliveira et al., 2016). Both extractive and abstractive summary either generic or query type. Expression of main theme based on query is called query-based summary whereas of the major content without any additional information of the documents is called generic summary (Mani, 1999; Wan, 2010).
Based on dimension, documents to be consider for summarization task, can divided the problem into single or MDS (Fattah, 2009; Rautray, Balabantaray, & Bhardwaj, 2015). Generating summary from a document or document set is called single or multi document summarization respectively. As document set includes many similar or distinct documents, therefore MDS is considered as extension of single document summarization. Due to large space in MDS, it is more critical task to extract relevant sentences. Thus MDS is recognized as an optimization problem. The main objective of MDS is to produce optimal informative summary of the original contents. However, swarm-based optimization techniques are the advisable options to address this optimization problem. In recent past, various meta-heuristic techniques such as particle swarm optimization (PSO) (Binwahlan, Salim, & Suanmali, 2009; Alguliev, Aliguliyev & Mehdiyev, 2011; Alguliev et al., 2011; Asgari, Masoumi, & Sheijani, 2014; Rautray, Balabantaray & Bhardwaj, 2015; Rautray & Balabantaray, 2015), differential evolution (DE) (Aliguliyev, 2009; Alguliev, Aliguliyev & Mehdiyev, 2011; Alguliev, Aliguliyev, & Hajirahimova, 2012; Alguliev, Aliguliyev, & Isazade, 2012; Alguliev, Aliguliyev & Hajirahimova, 2012; Alguliev, Aliguliyev & Isazade, 2013; Nandhini & Balasundaram, 2014), harmonic search (HS) (Shareghi & Hassanabadi, 2008), ant colony optimization (ACO) (Mosa, Anwar, & Hamouda, 2018; Hassan, 2015) cuckoo search (CS) (Mirshojaei & Masoomi, 2015) and genetic algorithm (GA) (Gordon, 1988; López‐Pujalte, Guerrero‐Bote, & de Moya‐Anegón, 2003; García, de Moya Anegón, & Zarco, 2000; Alguliev & Aliguliyev, 2005; Fattah & Ren, 2009; He et al., 2006; Zhao & Tang, 2010; Kogilavani & Balasubramanie, 2010) are applied both in single and multi-document summarization. From the different applications of cuckoo search algorithm, the author has inspired and presented cuckoo search algorithm-based summary extractor. Further the model is also compared with PSOE, IPSOE, CSOE and ACOE. Performance of each summary generated by different models is analyzed in terms of sentence-sentence similarity, ROUGE score, and readability metric. The experimental result is analyzed over DUC (Document Understanding Conference) datasets and, it is clearly observed that the performance of CSMDSE is showing significant result than PSO, IPSO, ACO and CSO based summary extractor.