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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Counterfactual Data Augmentation (CDA) is a commonly used technique forimproving robustness in natural language classifiers. However, one fundamentalchallenge is how to discover meaningful counterfactuals and efficiently labelthem, with minimal human labeling cost. Most existing methods either completelyrely on human-annotated labels, an expensive process which limits the scale ofcounterfactual data, or implicitly assume label invariance, which may misleadthe model with incorrect labels. In this paper, we present a novel frameworkthat utilizes counterfactual generative models to generate a large number ofdiverse counterfactuals by actively sampling from regions of uncertainty, andthen automatically label them with a learned pairwise classifier. Our keyinsight is that we can more correctly label the generated counterfactuals bytraining a pairwise classifier that interpolates the relationship between theoriginal example and the counterfactual. We demonstrate that with a smallamount of human-annotated counterfactual data (10%), we can generate acounterfactual augmentation dataset with learned labels, that provides an18-20% improvement in robustness and a 14-21% reduction in errors on 6out-of-domain datasets, comparable to that of a fully human-annotatedcounterfactual dataset for both sentiment classification and questionparaphrase tasks.

Improving Classifier Robustness through Active Generation of Pairwise Counterfactuals