English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Saliency maps can explain a neural model's predictions by identifyingimportant input features. They are difficult to interpret for laypeople,especially for instances with many features. In order to make them moreaccessible, we formalize the underexplored task of translating saliency mapsinto natural language and compare methods that address two key challenges ofthis approach -- what and how to verbalize. In both automatic and humanevaluation setups, using token-level attributions from text classificationtasks, we compare two novel methods (search-based and instruction-basedverbalizations) against conventional feature importance representations(heatmap visualizations and extractive rationales), measuring simulatability,faithfulness, helpfulness and ease of understanding. Instructing GPT-3.5 togenerate saliency map verbalizations yields plausible explanations whichinclude associations, abstractive summarization and commonsense reasoning,achieving by far the highest human ratings, but they are not faithfullycapturing numeric information and are inconsistent in their interpretation ofthe task. In comparison, our search-based, model-free verbalization approachefficiently completes templated verbalizations, is faithful by design, butfalls short in helpfulness and simulatability. Our results suggest thatsaliency map verbalization makes feature attribution explanations morecomprehensible and less cognitively challenging to humans than conventionalrepresentations.

Saliency Map Verbalization: Comparing Feature Importance Representations from Model-free and Instruction-based Methods