Conference Review: Information Foraging and Retrieval Autumn School

Following the tradition of past successful Information Foraging Summer Schools and the German Information Retrieval Autumn School, the Autumn School 2014 for Information Foraging and Retrieval took place from September 21-26, 2014, at Schloss Dagstuhl in Germany. It was co-organized by the German Information Retrieval Specialist Group and Radboud University in Nijmegen, The Netherlands. The school addressed students, early career researchers and practitioners in interactive information retrieval. The aim was to introduce students to advanced theory, models and technology related to information retrieval, interaction with information and information seeking as well as to discuss current research in these areas. The school attracted Master and PhD students in computer science and information science and practitioners from information infrastructure institutions. Overall, 27 students from over 10 countries participated; 13 of them benefited from a grant which was possible thanks to generous support by ACM SIGIR, the European Science Foundation – Project Elias, the BCS IRSG and the German Gesellschaft für Informatik – Fachgruppe Information Retrieval.

The notorious Dagstuhl group photo.
The notorious Dagstuhl group photo.

The school consisted of one week of lectures, which were delivered by experts in information retrieval and information foraging. The social program included networking in the Dagstuhl castle and a trip the historic city of Trier, which encompassed a guided tour.

 

The first lecture on IR Models by Ingo Frommholz introduced basic notions and the mathematical background of common IR models at a very high level. The difficulties arising from different aspects like document lengths etc. were discussed as well as the assumptions the formal models are based on.

A second foundational lecture on evaluation by Thomas Mandl reviewed the assumptions made by the traditional evaluation paradigms in IR. The Cranfield model and its weaknesses and strengths were discussed. The presentation included an overview of evaluation initiatives and an outlook into user oriented evaluation.

Applied Mathematics
Applied Mathematics

The talk on Activity Logging for Context and User Modeling by Wessel Kraaij introduced the topic of deeper analysis of user behavior based on logging users. The lecture explained how such data can be applied to improve systems, for example, in recommender systems.

 

In her session on Task Based Evaluation, Pia Borlund extended the perspective of evaluation into studies involving users. The problems and specific issues of organizing, conducting and analyzing a good study were thoroughly discussed.

In her Information Behavior lecture, Katriina Byström gave an overview over the well established models of Information Behavior. The lecture broadened the topic to include realistic models of work tasks and how they can be designed with information seeking contexts.

The second lecture on IR models by Thomas Rölleke presented more advanced concepts. The lecture went on to explain differences by comparing several models and by reaching a unified perspective when looking at them.

The session Modelling Interactive IR by Norbert Fuhr focused on three aspects of interactive retrieval, namely quantitative modeling, cognitive models, and user interface design. For the quantitative models, the interactive probability ranking principle was introduced along with methods for estimating the required parameters and for constructing Markov models of the user’s interaction with the system.

And another group photo with different lecturers.
And another group photo with different lecturers. Note the slight Moiré effect on Norbert Fuhr

The lecture Multimedia Retrieval by Stefan Rueger moved away from text retrieval and showed how content based image retrieval can be implemented today. The lecture deeply explained how features can be extracted from images and how they can be exploited for retrieval.

 

A happy face :-)
A happy face 🙂

The final talk of the autumn school was delivered by Ralf Schenkel on Efficiency Issues. Ralf focused first on efficient query processing and the necessary data structures for modern IR technology. Then he introduced MapReduce and showed how queries can be processed in a distributed manner.

 

The autumn school was an exciting event for students and lectures alike. Students had the opportunity to present their research topic in a special poster session, in which they were able to get in touch with senior experts of their field. Discussions were continued during an excursion to the historic city of Trier and of course in the evening over a cheese platter in the Dagstuhl wine cellar. All in all the Autumn School was a huge success and plans are underway for a follow-up school in 2016 at Schloss Dagstuhl.

Thanks to Mustafa Ghulam for providing the photos. An earlier version of this report was created by Thomas Mandl.