English

Interpretation of NLP models through input marginalization

Computation and Language 2020-10-28 v1 Artificial Intelligence

Abstract

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.

Keywords

Cite

@article{arxiv.2010.13984,
  title  = {Interpretation of NLP models through input marginalization},
  author = {Siwon Kim and Jihun Yi and Eunji Kim and Sungroh Yoon},
  journal= {arXiv preprint arXiv:2010.13984},
  year   = {2020}
}

Comments

10 pages, 5 figures, to be published in the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020)

R2 v1 2026-06-23T19:40:17.453Z