anserini-indexing

Tools to build Anserini (Pyserini) indices and evaluate document retrieval for (mainly) fact-checking datasets.

Install

Create Python virtual environment

The following commands load needed modules on RCI cluster. Install similar versions in your environment. See Pyserini instructions.

module load binutils/2.32-GCCcore-8.3.0
module load IPython/7.9.0-fosscuda-2019b-Python-3.7.4
module load GCCcore/9.3.0
module load Java/11.0.2

Create the environment and install requirements

python -m venv anserini_venv

source anserini_venv/bin/activate

pip install -r requirements.txt

Building indices & Evaluation

See slurm directory for SLURM scripts which should be run in order for FEVER and CTK datasets. Note that the SLURM scripts are plain bash scripts with additional information on hardware resources defined in header comments.