Summarization of related work through citations

Research output: Chapter in Book/Report/Conference proceedingConference contribution

View graph of relations Save citation



Automatically generating the related work section for a writing paper is useful for researchers as it can save a lot of time and avoid missing related works. The related work section of scientific paper usually introduces other researchers' work published before and makes comparisons with the current author's work. This paper proposes an approach to automatically generating related work by comparing the main text of the writing paper with the citations to the references from other papers. Our approach firstly collects the papers that cite the reference papers of the writing paper and extracts the corresponding citation sentences to form a citation document. It then makes a summarization that makes a comparison between the citation document and the writing paper's abstract, introduction and conclusion. It extracts the representative keywords from the citation document and the writing paper, and constructs a graph of the keywords. The discriminated nodes in the graph are figured out, and then the minimum Steiner tree that covers the nodes is extracted. A summary is generated by extracting the sentences covering the Steiner tree. The experiments show that our approach outperforms three baselines MEAD, ReWoS and ARWG according to ROUGE evaluation results.

Request a copy

Request a copy


Publication date11 Jan 2017
Publication titleProceedings - 2016 12th International Conference on Semantics, Knowledge and Grids, SKG 2016
EditorsHai Zhuge, Xiaoping Sun
Number of pages8
ISBN (Electronic)978-1-5090-4795-6
ISBN (Print)Piscataway, NJ (US)
Original languageEnglish
Event12th International Conference on Semantics, Knowledge and Grids, SKG 2016 - Beijing, China


Conference12th International Conference on Semantics, Knowledge and Grids, SKG 2016

Bibliographic note



  • citation, related work, summarization


Research outputs

Employable Graduates; Exploitable Research

Copy the text from this field...