Gianmaria Silvello

You are here

31 Oct 2016

Gianmaria Silvello

Associate professor/researcher at the Department of Information Engineering of the University of Padua

Gianmaria Silvello is associate professor/researcher at the Department of Information Engineering of the University of Padua. He was awarded a master degree in Computer Engineering, University of Padua in 2006, a post-graduate master on “Design, Management and Conservation of Public and Private Digital Archives” in 2007 and a Ph.D. in Information Engineering from University of Padua in 2011. He has published more than 80 scientific papers on data citation; data models; IR evaluation; and, digital libraries and archives. More about his research at


When: Day 2 - 15th November, Session 9 : DMP Technical Services #Part 2, 12:00 - 13:00

Reproducibility for IR evaluation

Abstract. Recently, a great deal of attention has been dedicated to the nature of research data and how to describe, share, cite, and reuse them to enable reproducibility in science and to ease the creation of advanced services based on them. Hence, reproducibility in Information Retrieval (IR) is rapidly becoming a key issue in the field, as witnessed, for instance, by the European Conference in IR (ECIR) series that from 2015 has allocated a whole paper track on reproducibility.

In this talk we present some methods and techniques for (semi-) automatically maintaining and enriching IR data and providing advanced services on top of them in order to enable reproducibility of experiments in the field. In particular, we will present the DIRECT system -  an open infrastructure able to manage and curate IR data - and LOD-DIRECT - a system to expose IR data as LOD in order to increase their potential, visibility and discoverability on the Web.

Since data citation is foundational for enabling reproducibility in IR, we also present some state-of-the-art methods we designed to automatically create human- and machine-readable citations to the data managed by DIRECT (i.e. citations of relational data) and exposed by LOD-DIRECT (i.e. citations to RDF data).

Reproducibility for IR evaluation from Research Data Alliance