In the NERC Study, a preliminary comparison of morphosyntactic
description schemes in corpus tagging has been carried out. The
underlying principle in view of a comparative assessment of the
facts covered by the individual tagging schemes is to allow for
different levels of granularity in the description and in the
respective tagging. For example, part-of-speech tagging may be
refined with morphosyntactic annotations, and one approach to
tagging might restrict itself to a subset of the inventory of
morphosyntactic features which are morphologically marked in a given
language, whereas another one might try to be exhaustive with
respect to morphologically marked and thus corpus-observable
differences. The fact that morphosyntax for a given language is
mostly a problem of different classifications of a finite domain
allows for the comparison of classification schemes (via their
tagging schemes) by use of the variable-granularity approach.
A different methodology has been chosen in NERC for the
syntactic area. There, the problem is to be able to relate complex
syntactic descriptions, for example from the functional level, with
their corpus-observable (and thus intersubjectively reproducible)
categorial and morphosyntactic reflexions. Part of this work will be
carried out in the DELIS project, where a number of exemplary verb
classes will be syntactically and semantically described and
descriptions will be related with corpus-searchable evidence.