[developers] TERG named entity regression?

Stephan Oepen oe at ifi.uio.no
Wed May 20 13:21:21 CEST 2009

hi eric,

> I think I have found a reversion in the latest LOGON TERG.


> Are you trying something new with named rel generation, perhaps?
> At Francis' suggestion, I reverted to the 0902 TERG release, and
> my problems were solved.

you are right, the current ERG `trunk' has broken the generation of
`unknown' names (it appears to work for a `named_unk_rel', but that
should not be part of the external interface, i.e. generator inputs
should of course always use `named_rel', whether a particular name
occurs in the ERG lexicon or not).  i think this is related to the
more recent email from francis about generating with generics, and
i will reply separately in that thread.

in general, i am really happy you are experimenting with more recent
versions of the ERG for LOGON-based MT (that is what i want to make
possible through the `terg' stub directory in the LOGON tree).  as a
rule of thumb, however, i would recommend you go with tagged release
versions, i.e. right now the 0902 release.

dan and i are preparing a new release (in time for barcelona), where
we frequently and liberally check in revisions to the `trunk'.  thus
you are better off working with the stable 0902 just now.

i will also propose a new snapshot of the complete LOGON tree, where it
would be desirable to have the various MT language pairs use the latest
ERG for parsing or genering english.  more on that on the `logon' list.

                                                         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