Paper: Extractive vs. NLG-based Abstractive Summarization of Evaluative Text: The Effect of Corpus Controversiality

ACL ID W08-1106
Title Extractive vs. NLG-based Abstractive Summarization of Evaluative Text: The Effect of Corpus Controversiality
Venue International Conference on Natural Language Generation
Session Main Conference
Year 2008
Authors

ve Summarization of Evaluative Text: The Effect of Corpus Controversiality Giuseppe Carenini and Jackie Chi Kit Cheung1 Department of Computer Science University of British Columbia Vancouver, B.C. V6T 1Z4, Canada {carenini,cckitpw}@cs.ubc.ca Abstract Extractive summarization is the strategy of concatenating extracts taken from a corpus into a summary, while abstractive summariza- tion involves paraphrasing the corpus using novel sentences. We define a novel measure of corpus controversiality of opinions con- tained in evaluative text, and report the results of a user study comparing extractive and NLG-based abstractive summarization at dif- ferent levels of controversiality. While the ab- stractive summarizer performs better overall, the results suggest that the margin by which ...