Interpretation of NLP models through input marginalization
This work addresses the out-of-distribution issue in interpreting NLP models, which is an incremental improvement for researchers and practitioners in explainable AI.
The authors tackled the problem of misleading interpretations in NLP models caused by out-of-distribution inputs from existing token-erasure methods, and proposed input marginalization as a remedy, demonstrating its application on sentiment analysis and natural language inference models.
To demystify the "black box" property of deep neural networks for natural language processing (NLP), several methods have been proposed to interpret their predictions by measuring the change in prediction probability after erasing each token of an input. Since existing methods replace each token with a predefined value (i.e., zero), the resulting sentence lies out of the training data distribution, yielding misleading interpretations. In this study, we raise the out-of-distribution problem induced by the existing interpretation methods and present a remedy; we propose to marginalize each token out. We interpret various NLP models trained for sentiment analysis and natural language inference using the proposed method.