README.md

this-before-that

Causal precedence relations in the biomedical domain

Getting the corpus (and other resources)

The annotated corpus can be found here. The word embeddings can be found here.

Running the sieve-based architecture (sans-lstm)

What you'll need...

1. Java 8
2. sbt

Rules used in the deterministic models

Three sets of Odin-style rules were used for the deterministic models:

1. Inter-sentential patterns
2. Intra-sentential patterns
3. Precedence markers
4. Reichenbach rules for tense and aspect

Running the LSTM

Installation: Using conda and Python 3.X

1. Fork and clone this repository

2. Install conda

3. Create a new conda environment using the environment.yml config:

conda env create -f environment.yml

The environment can be updated using the following command:

conda env update -f environment.yml

3. Activate the environment:

source activate bionlp

4. Test the installation:

python -c "import keras; print('Keras version: ', keras.__version__)"

GPU training

If you're training on an Ubuntu system with a CUDA card, you can run gpu_dependencies.sh to set things up.

Running the notebooks

jupyter notebook