GOFAIsum: a symbolic summarizer for DUC (2007)
| Venue: | In Proc. of DUC-07 |
| Citations: | 2 - 0 self |
BibTeX
@INPROCEEDINGS{Gotti07gofaisum:a,
author = {Fabrizio Gotti and Guy Lapalme and Montréal Québec and Luka Nerima and Eric Wehrli and Université De Genève},
title = {GOFAIsum: a symbolic summarizer for DUC},
booktitle = {In Proc. of DUC-07},
year = {2007}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
This article presents GOFAISUM, a topicanswering and summarizing system developed for the main task of DUC 2007 by the Université de Montréal and the Université de Genève. We chose to use an all-symbolic approach, the only source of linguistic knowledge being FIPS, a multilingual syntactic parser. We further attempted to innovate by using XML and XSLT to both represent FIPS’s analysis trees and to manipulate them to create summaries. We relied on tf·idf-like scores to ensure relevance to the topic and on syntactic pruning to achieve conciseness and grammaticality. NIST evaluation metrics show that our system performs well with respect to summary responsiveness and linguistic quality, but could be improved to reduce redundancy. 1
Citations
No citations identified.
