Let’s Summarize Scientific Documents! A Clustering-Based Approach via Citation Context

Scientific documents are getting published at expanding rates and create challenges for the researchers to keep themselves up to date with the new developments. Scientific document summarization solves this problem by providing summaries of essential facts and findings. We propose a novel extractive summarization technique for generating a summary of scientific documents after considering the citation context. The proposed method extracts the scientific document’s relevant sentences with respect to citation text in semantic space by utilizing the word mover’s distance (WMD); further, it clusters the extracted sentences. Moreover, it assigns a rank to cluster of sentences based on different aspects like similarity with the title of the paper, position of the sentence, length of the sentence, and maximum marginal relevance. Finally, sentences are selected from different clusters based on their ranks to form the summary. We conduct our experiments on CL-SciSumm 2016 and CL-SciSumm 2017 data sets. The obtained results are compared with the state-of-the-art techniques. Evaluation results show that our method outperforms others in terms of ROUGE-2, ROUGE-3, and ROUGE-SU4 scores.