This has been primarily used to visualize the importance scores of different types of
elements in XAI learning systems, such as showing input-output word alignment, highlighting words in input text or displaying extracted relations. 

Saliency-based visualizations are popular because they present visually perceptive
explanations, and they can be easily understood by different types of end users.

University of Michigan - Ann Arbor

The major visualization techniques include:

- Saliency

- Raw declarative representations

- Natural language explanation

XAI Visualization Techniques

Danilevsky, M., Qian, K., Aharonov, R., Katsis, Y., Kawas, B., & Sen, P. (2020). *A survey of the state of explainable AI for natural language processing*. arXiv preprint arXiv:2010.00711.

A Survey of the State of Explainable AI for Natural Language Processing

Visualization technique: Saliency

This visualization technique directly presents the learned declarative representations,
such as logic rules, trees, and programs. Such techniques assume that end users can
understand specific representations, such as first order logic rules and reasoning trees, and therefore may implicitly target more advanced users.

Raw declarative representations

It is verbalized in human-comprehensible natural language. The natural language can be generated using sophisticated deep learning models, or by using simple template based approaches.

Learn Before

Related