Andrew
Trotman and Shlomo Geva
At INEX there is
an underlying assumption that XML-retrieval and element retrieval are one and
the same. This is, in fact, not the case. The hypothesis at INEX is that XML
markup is useful for information retrieval. We firmly believe this, but no
longer in element retrieval. In this contribution we examine in detail the
evidence collected in support of element retrieval and suggest that, contrary
to expectation, it in fact supports passage retrieval and not element
retrieval. Particularly, we draw on other studies that collectively show that
INEX assessors are identifying relevant passages (not elements), they agree on
where in a document those passages lie, that there already exists suitable
metrics in the XML-retrieval community for evaluating passage retrieval
algorithms, and that the tasks make more sense as passage retrieval tasks.
Finally we show that future tasks of XML-retrieval also fit well with passage
retrieval.