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).

Dependencies

• Java 8
  • sudo add-apt-repository ppa:webupd8team/java
  • sudo apt-get update
  • sudo apt-get install oracle-java8-installer
• Maven
  • sudo apt-get install maven
• Latest version of Bazel
  • See installation instructions here: https://bazel.build/versions/master/docs/install-ubuntu.html
• Boost
  • sudo apt-get install libboost-all-dev
  • Note that boost is not directly used, but it is a dependency for legacy reasons.

Setting Up

• Make user the JAVA_HOME environment variable is set correctly, e.g.
  • export JAVA_HOME=/usr/lib/jvm/java-8-oracle
• Run ./setup.sh to download data and compile JNI binaries.
• Download and extract CCGBank data from https://catalog.ldc.upenn.edu/LDC2005T13.
• Move the ccgbank_1_1 directory to the data directory.

Running Experiments

• The experiments directory contains .conf files that specify experiments and stages of the experiment.
• Use ./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. localhost:8080.

Learning

• Training: ./run.sh experiments/train.conf train 8080
• Dev evaluation: ./run.sh experiments/eval.conf eval-checkpoints 8081

Demo

• Demo with the released model: ./run.sh experiments/demo.conf demo 8080
• Go the experiment summary page to access the demo.