Please use this identifier to cite or link to this item:
|Title:||Multidocument summarization of engineering papers based on macro- and microstructure|
|Keywords:||document structure analysis|
|Source:||Zhan, J., Liu, Y., Loh, H.T. (2011). Multidocument summarization of engineering papers based on macro- and microstructure. Journal of Computing and Information Science in Engineering 11 (1) : -. ScholarBank@NUS Repository. https://doi.org/10.1115/1.3563048|
|Abstract:||This paper focuses on automatic summarization of multiple engineering papers. A summarization approach based on documents' macro- and microstructure has been proposed. The macrostructure consists of a list of ranked topics from engineering papers. Topics are discovered by extracting and grouping frequently appearing word sequences into equivalence classes. Hence, the macrostructure symbolically presents the topical links in different papers. Meanwhile, the microstructure is defined as the rhetorical structure within a single paper. The identification of microstructure is approached as a classification problem. Each sentence in a paper is automatically labeled with one of the predefined rhetorical categories. Unlike existing summarization methods that first separate documents into nonoverlapping clusters and then summarize each cluster individually, our approach aims to summarize multiple documents according to the characteristics suggested at macro- and microstructure levels. The experimental study showed that our proposed approach outperformed peer systems in terms of recall-oriented understudy for gisting evaluation scores and readers' responsiveness. In an independent manual categorization task using the summaries generated by our approach and peer systems, we also performed better in terms of precision and recall. © 2011 American Society of Mechanical Engineers.|
|Source Title:||Journal of Computing and Information Science in Engineering|
|Appears in Collections:||Staff Publications|
Show full item record
Files in This Item:
There are no files associated with this item.
checked on Dec 9, 2017
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.