Self-Attention Network (SANet)

This repository contains an implementation of SANet (Self-Attention Network) presented in “Importance of Self-Attention for Sentiment Analysis“ by Letarte, Paradis, Giguère and Laviolette, presented at the EMNLP 2018 workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP.

Requirements

• Python 3.6
• Numpy 1.17.2
• Pytorch 1.3.1
• Poutyne 0.7.2
• Scikit-learn 0.22.2
• Pandas 0.25.1
• Nltk 3.4.1

BiBTeX

@inproceedings{letarte2018importance,
  title={Importance of self-attention for sentiment analysis},
  author={Letarte, Ga{\"e}l and Paradis, Fr{\'e}d{\'e}rik and Gigu{\`e}re, Philippe and Laviolette, Fran{\c{c}}ois},
  booktitle={Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP},
  pages={267--275},
  year={2018}
}