Within the context of the LS-GRAM project
an investigation has been done on the
reusability of LFG-based grammar
resources for the ALEP framework.
The LFG-grammar for German taken as the starting point of this investigation has
been developed at the Institut für maschinelle Sprachverarbeitung
(IMS) at the University of Stuttgart (see [Berman 1995]) and
implemented within the Grammar Work Bench of
Xerox (GWB). Our work was also based on the results of a study on the (non-)equivalences of the distinct formal devices implemented
in ALEP and other unification-based formalisms (see
[ET-10/52-93]). But our investigation was rather
concerned with a real experiment, achieving an output
equivalence between the source LFG-grammar and the target ALEP-grammar.
In this paper I will show how it has been taken
advantage of some structural identities of LFG and ALEP\
(both being a lexical-driven formalism combined with an obligatory context-free
backbone) in order to migrate the lexicon and structure rules.
More work was needed in order to migrate the feature descriptions of
the LFG grammar into ALEP (being a
typed formalism allowing a simple hierarchy). In this case, one
has to consider also the
completeness and coherence conditions put on the feature structures of
LFG and the distinct kinds of
(constraining) equations defined in the formalism. This information is enough for the definition of the type system of the
ALEP grammar. ALEP being a so-called ``lean''
formalism, it was not possible to
straightforward reproduce the compact
LFG-formulation of rules and lexicon entries.
I will also present some work recently done within the ALEP platform which led us into the field
of grammar engineering.
This work is dedicated to the integration of an external morphology and part-of-speech information provided
by preprocessing tools and taggers into ALEP. These experiments have led to a
reformulation of the linguistic resources defined within the ALEP\
grammars, allowing a very compact formulation (for example generic
lexicon entries and reduction of the set of syntactic rules), which
should facilitate the migration of ALEP resources to grammar
formalisms supporting richer formal devices.
The question which will remain: how can this approach been made
available for other platforms supporting unification-based grammars,
like for example the GWB or the XLE platform.