In HPSG, linearisation treatments for linguistic phenomena have been
advocated for some time now, and by several authors (e.g. Reape 1996,
Kathol 1995, Mueller 1999, Penn 1999). As a consequence of this,
proposals have emerged to process such grammars and parse with them.
The internal representation of a grammar has an important influence on
how effective parsing is. The simplest is probably where the grammar
is interpreted. A problem with such approaches is that processing
does not happen in an optimal way: it follows closely the form in
which the data are presented, and does not take advantage of known
properties of the parser and internal data structures. With
linearisation grammars, this effect is even stronger, because the
context-free backbone must be changed to deal with discontinuity.
While this example is obviously a simplification, it makes clear what
our aim is: compiling linearisation grammars into data structures that
allow efficient processing. In the talk, I shall present ongoing work
on such a parser, and associated compilation procedures, for
linearisation grammars. The grammars we work with are ALE grammars
which can be specified with linearisation constraints. Currently the
constraints are limited to a few basic ones: isolation, weak and
immediate precedence. The compilation procedure re-orders the
grammars such that linearisation information is moved as much as
possible into the control structure of the compiled form of the rules.
*******************************************************
Bitte den ausserordentlichen Termin beachten.
Fuer zusaetzliche Information ueber das CL-Kolloquium siehe auch
http://www.coli.uni-sb.de/~saurer/cl-kolloquium.html