Evaluating DUC 2004 Tasks with the QARLA Framework

摘要

This papers reports the application ofthe QARLA evaluation framework to theDUC 2004 testbed (tasks 2 and 5). Ourexperiment addresses two issues: howwell QARLA evaluation measures correlatewith human judgements, and what additionalinsights can be provided by theQARLA framework to the DUC evaluationexercises.