Browsing Information Retrieval System Results

Overview & Synopsis

Evaluation in information retrieval (IR) is crucial. Since the seventies, researchers use Cranfield or TREC-like framework to evaluate their systems and approaches resulting in system effectiveness calculated across reference collections. While numerical results are common practice for system comparison, we think that visual comparisons could also be of help for researchers. To this end, we developed an interface that allows IR scientists to compare system effectiveness. It relies on results from the trec_eval tool. At this stage, the interface allows analyses for ad hoc retrieval. This paper presents the interface.