Two is company — industry and academia in IR
Moderated by Mark Sanderson (RMIT)
Panel members: Behshad Behzadi (Google), Norbert Fuhr (University of Duisburg-Essen), Birger Larsen (Aalborg University), and Peter Mika (Yahoo!)
The relation between academic and industrial research in information retrieval has been a frequent topic of discussion.
In discussions about this relation, the use of proprietary data in IR research is one issue that gives rise to a range of positions. For instance, it was recently decided to turn ICTIR into a more ‘foundational’ conference, with an emphasis on experiments on data that is openly available to the research community. This move probably reflects the position that for scientific results to be reproducible, the data being used should be open. The inclusion of industry days or industry sessions at recent editions of SIGIR and ECIR represents a different position, namely that valuable insights can be gained from proprietary data.
But there’s more. As the community’s research agenda develops, a strong emphasis on online algorithms and online evaluation is emerging. Is this shift a source of divergence between academic and industrial research in IR? Work on online evaluation may require access to live systems. And it already seems that expertise on online algorithms and time-aware ranking methods can more easily be found in industrial research environments than in academia.
The ECIR 2014 organizers organize a panel to discuss the IR research ecosystem and the roles of academic and industrial research in this ecosystem. Our goal is to stimu- late discussion in the IR community that may inform future strategy development.