[developers] Re: Morphological ambiguity and xfst-lkb interface
Ann Copestake
Ann.Copestake at cl.cam.ac.uk
Sat Apr 16 04:29:29 CEST 2005
as threatened, I have started rewriting the tokenisation and morphology
components of the LKB. I have checked in the current code as a branch called
tok-and-morph. The active parser is currently broken and there are some more
minor infelicities. The revisions are described to some extent in comments in
parse.lsp, but the basic points are:
1) work towards allowing a full chart input from the tokeniser
2) work towards allow several different types of interface to morphology
3) clean up morphology and multiword code to use the chart in a reasonably
straightforward way (believed basically done). morph-history etc have gone,
as has *morphs* - this may break some of the more advanced user-fns
I am actively working on this and it's probably not worth testing yet, but if
you're about to develop any code that may interact, it would be worth checking.
I intend to check in a very artificial test grammar to check functionality
properly.
Ann
More information about the developers
mailing list