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

Spurious correlations threaten the validity of statistical classifiers. Whilemodel accuracy may appear high when the test data is from the same distributionas the training data, it can quickly degrade when the test distributionchanges. For example, it has been shown that classifiers perform poorly whenhumans make minor modifications to change the label of an example. One solutionto increase model reliability and generalizability is to identify causalassociations between features and classes. In this paper, we propose to train arobust text classifier by augmenting the training data with automaticallygenerated counterfactual data. We first identify likely causal features using astatistical matching approach. Next, we generate counterfactual samples for theoriginal training data by substituting causal features with their antonyms andthen assigning opposite labels to the counterfactual samples. Finally, wecombine the original data and counterfactual data to train a robust classifier.Experiments on two classification tasks show that a traditional classifiertrained on the original data does very poorly on human-generated counterfactualsamples (e.g., 10%-37% drop in accuracy). However, the classifier trained onthe combined data is more robust and performs well on both the original testdata and the counterfactual test data (e.g., 12%-25% increase in accuracycompared with the traditional classifier). Detailed analysis shows that therobust classifier makes meaningful and trustworthy predictions by emphasizingcausal features and de-emphasizing non-causal features.

Robustness to Spurious Correlations in Text Classification via Automatically Generated Counterfactuals