README.md

This split was used in the ACL 2014 paper. It uses only the proxy data. make_splits.sh will make the splits.

Test anything in year 2008.

Dev is anything in year 2007.

Train is everything else.

> grep '::preferred' amr-release-proxy.txt | sed 's/.*\(::id [^ ]*\).*\(::annotator [^ ]*\).*/\1 \2/' | sort | grep '_2008' | wc -l
2132
> grep '::preferred' amr-release-proxy.txt | sed 's/.*\(::id [^ ]*\).*\(::annotator [^ ]*\).*/\1 \2/' | sort | grep '_2007' | wc -l
2132
> grep '::preferred' amr-release-proxy.txt | sed 's/.*\(::id [^ ]*\).*\(::annotator [^ ]*\).*/\1 \2/' | sort | grep -v '_200[78]' | wc -l
3955