Generic document summarization approach based on controlled stochastic sentence selection

Abstract

In the new norm and cloud world era, online document generation has exponentially increased. The readers from different genres are unable to filter redundant information at a fast-paced rate. The research work is beneficial in raising awareness of utilizing online text summarization for distance learning among teachers, researchers, and students. It enables academia to quickly access concise and precise information from varied online sources. An efficient document summarization model reduces the read-time and improves information diversity; the paper presents an extractive summarization technique with a controlled stochastic sentence selection mechanism. The controlled stochastic limit is fine-tuned using TF, cosine, and Jaccard similarity measures. This unique sentence selection strategy is combined with a meta-heuristic approach to generate multiple solutions iteratively. The fitness of summary solutions is evaluated concerning the original document set producing the final summary. The various algorithms used for summarization are compared with the recommended model. The ROUGE-1 and ROUGE-2 values are empirically evaluated over DUC 2001, DUC 2002 datasets, which showcase an increase of 34.49% in Recall over the existing methods.