Recently, a technique called Layer-wise Relevance Propagation (LRP) was shown to deliver insightful explanations in the form of input space relevances for understanding feed-forward neural network classification decisions. In the present work, we extend the usage of LRP to recurrent neural networks. We propose a specific propagation rule applicable to multiplicative connections as they arise in recurrent network architectures such as LSTMs and GRUs. We apply our technique to a word-based bi-directional LSTM model on a five-class sentiment prediction task, and evaluate the resulting LRP relevances both qualitatively and quantitatively, obtaining better results than a gradient-based related method which was used in previous work.
@article{arxiv.1706.07206,
title = {Explaining Recurrent Neural Network Predictions in Sentiment Analysis},
author = {Leila Arras and Grégoire Montavon and Klaus-Robert Müller and Wojciech Samek},
journal= {arXiv preprint arXiv:1706.07206},
year = {2017}
}
Comments
9 pages, 4 figures, accepted for EMNLP'17 Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA)