System for the CLTL participation in SemEval2015 task 13: multilingual all-words sense disambiguation and entity linking.
You will need just to clone the repository and run the
instal.sh script, which will automatically download and install:
- ixa-pipe-pos: pos-tagger developed by the IXA group of the basque country (https://github.com/ixa-ehu/ixa-pipe-pos)
- It Makes Sense (IMS): wrapper around the IMS system for WSD to allow the use of NAF files (https://github.com/rubenIzquierdo/it_makes_sense_WSD)
- dbpedia_ner: wrapper around the dbpedia spotlight NER and NED for working with NAF files (https://github.com/rubenIzquierdo/dbpedia_ner)
- predominantsense: wrapper around predominant sense algorithm (https://bitbucket.org/MartenPostma/predominantsense)
The version of the IMS system only works with java 1.6, so make sure you are using that version, you can check with
java -version. If you are
using any other version, you will need to modify the file
libs/it_makes_sense_WSD/ims/testPlain.sh and set the
java command to the exact
Change: java -mx2500m -cp $CLASSPATH sg.edu. by: /usr/lib/jvm/java-1.6.0-openjdk-amd64/bin/java -mx2500m -cp $CLASSPATH sg.edu.....
So basically these are the the only steps required:
git clone https://github.com/cltl/vua-wsd-sem2015 cd vua-wsd-sem2015 install.sh
To run the whole system, you will need to call to the script
run.sh. This script will automatically call to all the modules
in our pipeline for processing the original SemEval2015 files that can be found at the folder
It will create a new folder
data_en_naf, where all the NAF files created will be stored. Different files with different suffixes
represent intermediate results of the whole pipeline. After running
run.sh, 'run_experiment.sh' will run the experiment. The output can be found in evaluation/stats.
- Ruben Izquierdo Bevia
- Marten Postma
- Vrije University of Amsterdam
##License Sofware distributed under GPL.v3, see LICENSE file for details.