VPE Version 1.0 (August 2010)

This is version 1.0 of the verb phrase ellipsis annotation for the Wall Street Journal (WSJ) sections of the Penn Treebank (PTB version 3.0). The stand-off annotations are provided in 25 files named after the section numbers of the WSJ and come with the extension ".ann". Each occurrence of VPE found in the corpus is listed on a separate line. Information for a VPE is organised in columns, which are indicated by spaces, organised by the following schema:

<FILENAME> <EB> <EE> <AB> <AE> <TR> <STA> <STP> ....

where <FILENAME> is the raw filename as distributed with the PTB, <EB> and <EE> are character positions marking the beginning and end of the ellipsis, and <AB> and <AE> the character positions marking the start and end of the antecedent. Column <TR> holds the trigger type, <STA> the syntactic type of the antecedent, and <STP> source-target pattern of the ellipsis.

Johan Bos University of Groningen bos@meaningfactory.com

Jennifer Spenader University of Groningen j.spenader@ai.rug.com