hier-class

hier-class copied to clipboard

Hierarchical Neural Attention-based Classification

Code and Data for the paper A Hierarchical Neural Attention-based Text Classifier published as a short paper in EMNLP, 2018.

Note: This repository is no longer actively maintained. Due to unavailibility of saved test splits and model files, reproducibility of the exact numbers of the paper is not guranteed.

Dependencies

• Pytorch
• Pandas
• tqdm

Dataset

• Web of Science dataset - Drive link
• DBPedia dataset (curated using get_taxonomy.py)
  • Drive link
  • Train test split - Drive link

Running the code

• Navigate into the directory codes/app/
• Run python main.py --config_id <config name>

Experiment configs

To run, create an experiment config from the sample configs in config/ folder.

Questions

Please open an issue!

License

MIT