Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Enhance BioNLP Shared Task (ST) Event Extraction (EE) output with external resources
Tree: 3092366442

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
ext Ignore liblinear remenants

eepura, eeプラ or ee+

This is just a dummy README for now.

Data Pre-processing

This section is probably not interesting to anyone but left in for personal reference and for those that really want to replicate each conversion step up until the experimental scripts take over.

Negation and speculation annotations were stripped using:

#!/usr/bin/env bash
for f in `find res -name '*.a2'`
    grep -v -E $'^M[0-9]+\t(Speculation|Negation) ' $f > ${f}.strp

Filenames were corrected for the nesp files since for some reason they violated the original BioNLP ST naming conventions by containing spaces:

find res -name '* *.nesp' | while read f
    mv "$f" `echo $f | sed -e 's| |_|g'`

nesp files were converted into stand-off for visualisation and to fit into the framework using the script:

for f in `find res/ann -name '*.nesp' | sed -e 's|\.nesp$||g'`
    src/ ${f}.{nesp,txt,a1} -t 1000 -e 1000 \
        `if [ -f ${f}.a2 ]; then echo "-a ${f}.a2"; fi;` \
        > ${f}

Frequently Asked Questions (FAQ)

  • Q: Why the name eepura?
  • A: eepura was developed at the University of Tokyo and draws some inspiration from Japanese. In Japanese C++ can be written shiipurasupurasu (シープラスプラス) and due to the wonderful nature of the Japanese language purasu (プラス) naturally becomes pura (プラ) for short, then due to yet another wonderful aspect of Japanese it is not uncommon to drop repeated patterns in expressions (happens with Japanese onomatopoeitic expressions quite frequently, but that is a completely different story) so informally C++ is shiipura (シープラ). Stroustrup intended the C++ name to signify the extension of the C language, thus building a system that enhances the output of existing event extraction systems would be...

  • Q: Is this really a FAQ?

  • A: Not really, no, mostly it is just a bunch of questions that might be raised written while waiting for the model to be trained during experiments
Something went wrong with that request. Please try again.