Information retrieval research, at least as conceived by the SIGIR community, is fundamentally experimental in nature. As such, the presentation of results from controlled, reproducible experiments lies at the core of our work. Many reports follow the same general format: authors propose a new retrieval method, whose performance on some well-defined task is compared against a baseline. Authors also report results from alternative configurations, e.g., variations in parameters, turning off (ablation) of different components, etc. The presentation of experimental results forms an integral part of the conferences and journals that comprise the medium in which knowledge is disseminated.
/lp/association-for-computing-machinery/presentation-schemes-for-component-analysis-in-ir-experiments-OdJzjheSSe