K. VIJAY-SHANKER and DAVID WEIR Department of Computer and Information Science, University of
Delaware and School of Cognitive and Computing Science, University of Sussex Underspecification of trees has, until recently, either not been
used to specify constraints on the well-formedness of grammars,
or has been applied to theories that are not formulated in a completely
precise way (see [1], [2,3], and [4,5] for exceptions to this).
The potential use of tree descriptions provides insights into
the formulation of various strongly lexicalized grammar formalisms
including Lexicalized Tree-Adjoining Grammars and related formalisms.
In this course we will show how this has led to the development
of D-Tree Grammars [6] which was designed with the expressed purpose
of exploiting the use of partial tree descriptions. Underspecification has also been used in parsing [7,8,9]. Tree
descriptions can be used by a parser to postpone the need to make
commitments about the structure(s) being assigned to that portion
of the input seen so far. We will describe various approaches
to the parsing of lexicalized grammars that seek to exploit underspecified
descriptions at intermediate points in the parsing process.
EXPLOITING TREE DESCRIPTIONS IN STRONGLY LEXICALIZED GRAMMAR FORMALISMS AND PARSERS
vijay@cis.udel.edu and davidw@dogs.susx.ac.uk
None