Output details

Relevance-based abstraction identification: technique and evaluation

<09> This paper was an invited submission, selected as one of the 8 best papers presented at RE’2010 (from 153 submissions), and was one of the 5 finally accepted. The paper is a benchmark in two ways. First, the algorithm that forms its primary contribution significantly improves upon the performance of existing algorithms. Secondly, the paper solves the problem of finding a gold standard data set, which is a perennial problem in this type of work. The data set, and more generally the methodology used, is proposed benchmark data set against which future algorithms can be evaluated.