Empirical 2014 - First International Workshop - Empirical 2014

The Empirical 2014 workshop is a half-day workshop that will open the Extended Semantic Web Conference 2014 and seeks to gather people from all fields to work together to enhance our evaluation methodology. Please see the Call for Papers.
Statistical inference using hypothesis-testing methods is not commonly used to support empirical results reported in the Semantic Web literature. Thus, given the impact that these statistical methods may have on different domains, attendees may see the workshop as a venue to discuss the applicability of these methods in their current studies. Trivial use of hypothesis testing is not in the scope of the workshop. The primary focus, as will be detailed in the call for papers, is on statistical methods to evaluate assertions on performance of complex systems. Hitherto, the focus of the community has been on benchmarking. Currently, it suffers from several problems, e.g., lack of summary statistics, no hypothesis tests, no structured approach to control of experiments, oversimplification of the benchmark in face of complex problems, etc.
Topics to be included (not limited):
Foundations of hypothesis formulation: triangularization, common constructs in the Semantic Web (e.g., performance, completeness) and their validity, typical operationalizations, and quantifiability.
Statistical methods to demonstrate significance and levels of confidence of experimental studies on Semantic Web technologies and empirical results.
Experimental methodologies to design reproducible experimental studies in the context of Semantic Web technologies.
Critical examinations of assumptions inherent in statistical practices in Semantic Web research (e.g., assuming normal distribution of populations).
Limitations of existing statistical methods on empirical studies of Big Data.
Statistical techniques to analyze benchmarks.
Read more about Workshop invitation
Format of the workshop
The Empirical 2014 workshop brings several novel elements to the conference. We open by an invited talk on statistics, to ensure that everyone is equipped to understand and engage in the rest of the workshop. Then, we have the usual presentations, but end each session with a two-chair panel, which is open to everyone. Finally, we end the workshop with The Trial, where benchmarking stands accused of being unscientific. You may submit your views to the defence and the prosecutor, who will prepare procedures and may call you as a witness.