[developers] Fwd: New stable version of ERG: 1111
danf at stanford.edu
Mon Dec 5 18:23:06 CET 2011
Hi all -
I made this announcement on the ERG mailing list last week (though it may still be lost in the ether somewhere), and was reminded that I should also send it to the developers' list. Two additional related notes:
- Stephan is aiming to produce a new version of the WikiWoods parsed corpus within the next few weeks, using this 1111 version of the ERG.
- An emerging treebank of the corpus used for the Penn Treebank (one million words) is being annotated by the Computational Linguistics dept. in Saarbruecken (Yi, Valia, and students), and we plan to release some initial portion of this treebank early in 2012, updated to this 1111 version.
----- Forwarded Message -----
From: "Dan Flickinger" <danf at stanford.edu>
To: erg at delph-in.net
Cc: danf at stanford.edu
Sent: Wednesday, November 30, 2011 2:05:10 PM
Subject: New stable version of ERG: 1111
I am pleased to announce the release of a new stable version of the ERG, namely "1111". This year's version of the grammar contains some expansion of linguistic coverage, along with some additional treebanked profiles stored as usual in the subdirectory erg/tsdb/gold. I hope that any surprises in behavior will be pleasant ones, and of course I welcome feedback as always.
Particular thanks to Stephan Oepen, who has not only made further improvements to the chart mapping and preprocessing components in this version, but also invested considerable effort in rebuilding the maxent and PCFG models for the grammar, and then took several iterations with me to arrive at the `gold' profiles for the various treebanked data sets now being curated as the Redwoods corpus.
With best regards,
More information about the developers