conll2009 icon

TrEd Extension: conll2009 (version 1.18)

CoNLL 2009 Shared Task

Description:

This package contains a stylesheet for visualization of CoNLL-2009-ST data in TrEd, scripts (for Linux and similar systems) for converting from the CoNLL-2009-ST format to PML, and a TrEd annotation mode for predicate-argument annotation.

TrEd macro files:

Stylesheet files:

Resource files:

Script files:

NOTE: The conversion script works only on Unix and similar systems where basic UNIX tools like bash and sed are installed.

Usage

Run the bash script

conll2009/bin/conll2009-to-pml.sh

with CoNLL-2009-ST data files as arguments. When finished, open the created *.pml data in TrEd, using the stylesheet conll2009.

Handling bogus data

In conversion from PML to CoNLL-2009-ST format, if a node contains apreds pointers to a node with empty attribute pred, warning is issued and the column PRED is filled with a special value 'pml:-NO-PRED-', which is transformed back to an empty string during the reverse conversion. This is because we do not want to drop the pointer (although possibly bogus) and the CoNLL-2009-ST format provides no way to store a pointer to a predicate with empty value in the 'PRED'.

Moreover, the CoNLL-2009-ST data format uses a column named FILLPRED which is used to indicate the words that should be considered predicates (should have PRED filled). The annotation macros do not attempt to set FILLPRED according to whether PRED is filled, however the conversion script from PML to CoNLL-2009-ST format has a flag (--compute--filpred) which can be used to set FILLPRED=Y for exactly those nodes where PRED is set.

Authors:

The core script conll2pml was created by Petr Pajas. The latest version of the script can be downloaded here. The rest was done by Jan Štěpánek in 2009. TrEd stylesheet was created by Jan Štěpánek, macros were written by Petr Pajas.

Copyright 2009,2010 by Petr Pajas and Jan Štěpánek

Acknowledgement

Supported by the grants 1ET101120503 and MSM 0021620838.