I’m a senior researcher in domain-specific search, with extensive experience in Information Retrieval Evaluation. Together with collaborators at TU Wien and RSA DSc we make use of what we call “test collections” to test various hypotheses about information extraction and processing, about results delivered by search systems. The test collections come from different domains, and may contain complete documents describing inventions (i.e. patent documents), scientific articles, tweets, search logs, or images. Knowing what the data contains and having it well described, at best following the FAIR principles, is becoming more important as data understanding is a critical component of having a good IR system.

From my own personal experience, data that we work with is seldom described to such a level that using it is straightforward and easy – unless we are the creators of that data. For data sets we are authoring, we strive to describe them so that other researchers can re-use them easily, and we make them available on institutional websites.