TU expert collection


Richard Berendsen
r.w.berendsen ( AT ) uva.nl


1. Introduction
2. Getting the data
3. Collection statistics

1. Introduction

The TU expert collection is based on the Webwijs (“Webwise”) system developed at Tilburg University (TU) in the Netherlands. It is an update of the UvT expert collection. Webwijs is a publicly accessible database of TU employees involved in research or teaching. Currently, Webwijs contains information about over 1147 experts. For many of them a research description and a list of publications is availale. In addition, each expert can select expertise areas from a list of 2507 areas; they can suggest new areas that need to be approved by the Webwijs editor. Each area has a separate page that shows all experts associated with it and, if available, a list of related expertise areas. The TU Expert Collection was compiled in December 2008.

Webwijs is available in Dutch and English, and this bilinguality has been preserved in the collection. Most Dutch Webwijs pages have an English translation. As for expertise areas, not all Dutch areas have an English translation, but the reverse is true: all English areas have a Dutch equivalent.

2. Getting the data

If you would like to use these resources, please drop us a line, and we will make the collection available for you. If you publish results obtained using the resources made available here, please include the following citation:

On the Assessment of Expertise Profiles, R.Berendsen, K. Balog, T. Bogers, A. van den Bosch and M. de Rijke. In Journal of the American Society for Information Science and Technology, 2013., pp. XXX-YYY, 2013.

3. Collection statistics

The TU expert collection comes with two main sets of ground truth. One is an updated version of the self-selected profiles, similar to the ground truth used in the UvT collection. The other consists of judged system-generated profiles. The table below has histograms of the relevance assessments in both sets of ground truth: self-selected profiles (Top) and judged system-generated areas (Bottom). On the left experts are binned according to how many knowledge areas are in their profile. From the point of view of expert profiling, the leftmost histograms have queries binned according to how many relevant items there are for them. On the right areas are binned according to in how many profiles they are included. From the point of view of expert finding, the rightmost histograms have queries binned according to how many relevant items there are for them.





Distribution of queries over number of relevant items for expert profiling (left) and expert finding (right). The top graphs are based on the self selected areas. The bottom graphs are based on the judged system-generated areas