[developers] Punctuation and "-default-les" type mapping in PET/ERG

Stephan Oepen oe at ifi.uio.no
Sun Apr 13 14:25:32 CEST 2008

hi richard (and c.j.),

> Can one of the PET developers check if this is a reasonable account
> of what might be happening, and, ideally, fix it?

i am afraid i have no SMAF experience with PET, and i suspect the same
likely is true for most of the other current PET maintainers.  ben is
the relevant expert here (he contributed the SMAF code in PET ).  ben,
are you currently in a position to support SMAF mode in PET?

--- in general, i find it desirable that we aim to reduce the various
mechanisms (in PET or the LKB) that frob feature structures behind the 
scenes, e.g. `stamping in' CARGs, manufacturing PREDs, and the like.

there is a forthcoming branch of PET that aims to transfer more control
into the grammar, in that it assumes that (a) all tokens are encoded as
(small) input feature structures (IFS) and (b) in lexical instantiation
the sequence of IFSs that trigger a lexical entry (LE), be it a regular
or generic entry, is the argument to the LE (i.e. lexical instantiation
is viewed as derivation over sequences of token IFSs).  in this setup,
the LE has access to the complete token information and can simply use 
re-entrancies to incorporate token-level values.  the general idea has
just been submitted to LREC, see:


we have started experimentation with the ERG in this framework, and our
hope is to (over time) arrive at a setup that is conceptually simpler.
this branch of PET is not ready for public use yet, i am afraid, but i
expect it should get to that point later this year, maybe even in time
for a demonstration and discussion at the DELPH-IN summit.

                                                      all best  -  oe

+++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
+++     CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
+++       --- oe at ifi.uio.no; oe at csli.stanford.edu; stephan at oepen.net ---

More information about the developers mailing list