Model-agnostic-methods for Text Classification

This repository contains the code to reproduce the results of the paper (Applying Model-agnostic Methods to Handle Inherent Noise in Large Scale Text Classification) accepted at COLING 2020.

Getting Started

Dependencies

• Python 3.7
• Pandas
• Numpy
• Keras
• SkLearn
• Pickle

Installing

git clone https://github.com/tayalkshitij/model-agnostic-methods.git
cd model-agnostic-methods/

How to use

Train main model:

python code/main_experiment/main.py <path_to_dataset>

Train noise model:

python code/noise_experiment/noise_main.py <path_to_dataset>

Dataset

Google drive link for the datasets are as follow:

Automotive Dataset Link. Beauty Dataset Link. Electronics Dataset Link.

Pre-trained Embeddings

Glove Link.

Author

If you have any question, please contact the author: Kshitij Tayal (tayal007@umn.edu)

License

See the LICENSE file for more details.

Citation

When using the dataset or code, please cite our paper:

@article{tayalmodel,
  title={Model-agnostic Methods for Text Classification with Inherent Noise},
  author={Tayal, Kshitij and Ghosh, Rahul and Kumar, Vipin},
  journal={The 28th International Conference on Computational Linguistics},
  year={2020}
}

Acknowledgements

The codebase is based off D2L and edufonseca. Both are great repositories - have a look!