<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> <style type="text/css" style="display:none;"></style> </head> <body dir="ltr"> <div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Helvetica,sans-serif;" dir="ltr"> Colleagues, I am pleased to announce the release of a new stable version of the English Resource Grammar, called ERG 2018.  The previous stable version "1214" has served us for long enough, and I hope that you will find the new version to be a useful upgrade in coverage and quality.  Please note that you will need up-to-date versions of the processing platforms in order to use the 2018 version, because it uses a recently adopted format for documentation strings in the TDL source code which is not backward compatible. To obtain a copy of the source for the new grammar, do the following svn checkout <a href="http://svn.delph-in.net/erg/tags/2018" class="OWAAutoLink" id="LPlnk564199" previewremoved="true"> http://svn.delph-in.net/erg/tags/2018</a> Appended to this message is a summary of the changes from 1214 to 2018.  I will be updating the wiki in various places to refer to the new grammar version, and hope that the online demo can soon also point to the new version. My thanks to the many among you who have contributed to this release, whether with code, data, suggestions, feature requests, or constructive critique.  Dan -------------------------------------------------------------- <div>Release notes for stable version "ERG 2018" Annotations - Supplied full-forest treebanks for Redwoods profiles, including the first five   sections of the WSJ. - Added profiles for WeSearch user-generated content (wlb03, wnb03), and for   Sherlock Holmes story (sh-spec). - Improved the well-formedness and consistency of the MRSs, aiming for more   consistency with an updated version of the semantic algebra. Token mapping - Upgraded to use GML 1.0, particularly relevant for the WeScience corpus. - Improved support for both `strong' brackets (manually inserted) and `weak'   delimiters (motivated by, for example, hyphens) to signal phrase boundaries   that should not be crossed. Syntax - Enabled extraction from within NPs. - Changed the attachment order of pre- and post-nominal modifiers, so now the   pre-modifiers attach before post-modifiers. - Added two new types of modifiers of verbal projections: indefinite NPs as in   |She walked out of the casino a rich woman.| and gapped clauses of saying, as   |They will, we suspect, leave early.| - Added the `do-be' construction as in |the only thing we said she had to do was   finish the assignment| Semantics - Moved information-structure constraints from RELS to ICONS in MRSs, including   for focus movement (`topicalization') and passivization. - Simplified the inventory of role labels, notably for conjunction relations,   where ARG1 and ARG2 replace the old L-HNDL, R-HNDL, L-INDEX, R-INDEX roles. - Improved SEM-I consistency. Platforms and applications - Added support for ubertagging with PET and ACE, thanks to Rebecca Dridan. - Added robust parsing mode for ACE using csaw PCFG, thanks to Woodley Packard. - Added support for `agree' parser/generator - Expanded the inventory of mal-rules for grammar-checking. - Added `transfer' rules and support for generation from first order logic. - Added support for manually inserted `strong' brackets which force phrasal   boundaries, as in |we ⌊(⌋ saw a man ⌊)⌋ with a telescope|. - Changed RNAME value on rules from string to type, to allow `weak' bracket   rules to constrain which rules must apply, as for named entities such as   |New York Stock Exchange|. - Added support for robust `bridging' rules (disabled by default). -----------------</div> </div> </body> </html>