Global Neural CCG Parsing with Optimality Guarantees
This repository contains the code for replicating results from Global Neural CCG Parsing with Optimality Guarantees at EMNLP (Lee et al., 2016).
- Java 8
sudo add-apt-repository ppa:webupd8team/java
sudo apt-get update
sudo apt-get install oracle-java8-installer
sudo apt-get install maven
- Latest version of Bazel
- See installation instructions here: https://bazel.build/versions/master/docs/install-ubuntu.html
sudo apt-get install libboost-all-dev
- Note that boost is not directly used, but it is a dependency for legacy reasons.
- Make user the
JAVA_HOMEenvironment variable is set correctly, e.g.
./setup.shto download data and compile JNI binaries.
- Download and extract CCGBank data from https://catalog.ldc.upenn.edu/LDC2005T13.
- Move the
ccgbank_1_1directory to the
.conffiles that specify experiments and stages of the experiment.
./run.sh <config> <goal> <port>to run an experiment that launches the specified goal stage and its dependent goal stages.
- An experiment summary (e.g. progress, logs, and intermediate results) is hosted locally, and is accessible via a web browser at the given port, e.g.
./run.sh experiments/train.conf train 8080
- Dev evaluation:
./run.sh experiments/eval.conf eval-checkpoints 8081
- Demo with the released model:
./run.sh experiments/demo.conf demo 8080
- Go the experiment summary page to access the demo.