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/

Zendy Plus

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

Zendy Tools

Zendy Open

A growing body of work shows that models exploit annotation artifacts toachieve state-of-the-art performance on standard crowdsourcedbenchmarks---datasets collected from crowdworkers to create an evaluationtask---while still failing on out-of-domain examples for the same task. Recentwork has explored the use of counterfactually-augmented data---data built byminimally editing a set of seed examples to yield counterfactual labels---toaugment training data associated with these benchmarks and build more robustclassifiers that generalize better. However, Khashabi et al. (2020) find thatthis type of augmentation yields little benefit on reading comprehension taskswhen controlling for dataset size and cost of collection. We build upon thiswork by using English natural language inference data to test modelgeneralization and robustness and find that models trained on acounterfactually-augmented SNLI dataset do not generalize better thanunaugmented datasets of similar size and that counterfactual augmentation canhurt performance, yielding models that are less robust to challenge examples.Counterfactual augmentation of natural language understanding data throughstandard crowdsourcing techniques does not appear to be an effective way ofcollecting training data and further innovation is required to make thisgeneral line of work viable.

Counterfactually-Augmented SNLI Training Data Does Not Yield Better Generalization Than Unaugmented Data