Back to Basics: CLASSY 2006

John M. Conroy, Judith D. Schlesinger, Dianne Prost O’Leary, Jade Goldstein · 2006

The IDA/CCS summarization system, CLASSY, underwent significant change for this year’s DUC. Two changes made processing simpler and faster: 1) we eliminated the use of a POS (part of speech) tagger for sentence splitting and to assist sentence trimming, and 2) we simplified the scoring of sentences for inclusion in the summary by introducing a new “approximate oracle” score. An additional change introduced a modest amount of extra computation: we ordered sentences in the summary using a new Traveling Salesperson (TSP) formulation. These changes improved ROUGE scores on the DUC 2005 data from last year and gave strong performance in the DUC 2006 competition.

Read the paper · More papers on PaperTik