Visualize: Figure1.
Description: "Example of the saliency scores for the words
(columns) of an instance from the Twitter Sentiment
Extraction dataset. They are produced by the explainability techniques (rows) given a Transformer model. The first row is the human annotation of the salient words."