Andrew
Trotman
Proceedings of the INEX 2005
Workshop on Element Retrieval Methodology pp. 58-64
Document centric
information retrieval is used every day by people all over the world. It is an application well studied, well
understood, and of which there is a sound user model. Element retrieval, on the other hand, is a
new field of research, with no identified applications, no users, and without a
user model.
Some of the
methodological issues in element retrieval are identified. The standard document collection (the INEX /
IEEE collection) is shown to be unsuitable for element retrieval, and the
question is raised – does such a suitable collection exist? Some characteristics of querying behavior are
identified, and the question raised – will users ever use structural hints in their
queries? Examining the judgments and
metrics, it is shown that the judgments are inconsistent and the metrics do not
measure the same things.
It is suggested
that identifying an application of element retrieval could resolve some of
these issues. Aspects of the application
could (and should) be modeled, resulting is a more sound field of element
retrieval. Alternatively, whatever it
is, users don’t want it, judges can’t judge it, and the metrics can’t measure
it.