search

alvations / disambiguate

Disambiguate is a tool for training and using state of the art neural WSD models
https://arxiv.org/abs/1905.05677
0 stars 0 forks source link

Installation #1

Open alvations opened 4 years ago

alvations commented 4 years ago 
sudo apt update
sudo apt install maven

pip install -U torch allenlp transfomers torchtext 
pip install -U configargparse sacrebleu

git clone https://github.com/getalp/UFSAC.git 
git clone https://github.com/goodmami/disambiguate.git

cd UFSAC/java
mvn install
cd ../..

cd disambiguate
./java/compile.sh
alvations commented 4 years ago

Is --data_path ufsac-public-2.1/ ?

alvations commented 4 years ago

At first look, we should have a Python version of the reader for https://github.com/getalp/UFSAC

That'll simplify maintenance and corpus reader for PyTorch.

Note: The synset mappings in *.xml are using the sensekeys, e.g. european%3:01:00::, NLTK default sense key mapping is rather broken, use wn instead.