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

A principle behind dozens of attribution methods is to take the predictiondifference between before-and-after an input feature (here, a token) is removedas its attribution. A popular Input Marginalization (IM) method (Kim et al.,2020) uses BERT to replace a token, yielding more plausible counterfactuals.While Kim et al. (2020) reported that IM is effective, we find this conclusionnot convincing as the DeletionBERT metric used in their paper is biased towardsIM. Importantly, this bias exists in Deletion-based metrics, includingInsertion, Sufficiency, and Comprehensiveness. Furthermore, our rigorousevaluation using 6 metrics and 3 datasets finds no evidence that IM is betterthan a Leave-One-Out (LOO) baseline. We find two reasons why IM is not betterthan LOO: (1) deleting a single word from the input only marginally reduces aclassifier's accuracy; and (2) a highly predictable word is always givennear-zero attribution, regardless of its true importance to the classifier. Incontrast, making LIME samples more natural via BERT consistently improves LIMEaccuracy under several ROAR metrics.

Double Trouble: How to not explain a text classifier's decisions using counterfactuals synthesized by masked language models?