Applying Layer-wise Relevance Propagation (LRP) to make Named Entity Recognition (NER) explainable. The NER model is implemented with Bi-directional LSTM (Bi-LSTM) and Conditional Random Field (CRF).
- Python 3.5
- Keras 2.1.5
- Sickit-learn 0.19.1
- Pandas 0.22.0
- Numpy 1.14.2
- Matplotlib 2.2.2
Explaining Recurrent Neural Network Predictions in Sentiment Analysis by Arras, Leila, et al, 2017
Explaining Recurrent Neural Network Predictions in Sentiment Analysis, code
Visualizing and Understanding Neural Models in NLP by J. Li, X. Chen, E. Hovy and D. Jurafsky, 2016
Visualizing and understanding neural machine translation by Ding, Yanzhuo, et al. 2017