The migration has been done manually, but a semi-automatically
migration between LFG and ALEP seems to be feasible (see the algorithm
described by Dieter Kohl and Stefan Momma in the ET 10/52 study, which
was mainly based on the Charon-System developed at the IMS in
Stuttgart, [ET-10/52-93]).
The decision to make this migration in a manual fashion ha been dictated
by two factors: the GWB platform was not transparent enough for this
kind of work, and we also needed the migrated grammar in a very short
time. So there was no time left for describing migration algorithms
between the GWB and the ALEP platforms
.
The migration has been done in two steps:
- porting of rules and migration of feature names onto ALEP type descriptions,
- precising the type system of ALEP, taking into account the filtering conditions of LFG and the kind of constraints used.
The work done so far has achieved an output equivalence between both grammars, having also an identical coverage of linguistic phenomena. This results have been obtained in a very short time (two days), taking into account that the person performing this migration had also to be introduced to the ALEP platform.