Interpretation of NLP models through input marginalization

Siwon Kim, Jihun Yi, Eunji Kim +1

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…