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

Despite pre-trained language models have proven useful for learninghigh-quality semantic representations, these models are still vulnerable tosimple perturbations. Recent works aimed to improve the robustness ofpre-trained models mainly focus on adversarial training from perturbed exampleswith similar semantics, neglecting the utilization of different or evenopposite semantics. Different from the image processing field, the text isdiscrete and few word substitutions can cause significant semantic changes. Tostudy the impact of semantics caused by small perturbations, we conduct aseries of pilot experiments and surprisingly find that adversarial training isuseless or even harmful for the model to detect these semantic changes. Toaddress this problem, we propose Contrastive Learning with semantIc NegativeExamples (CLINE), which constructs semantic negative examples unsupervised toimprove the robustness under semantically adversarial attacking. By comparingwith similar and opposite semantic examples, the model can effectively perceivethe semantic changes caused by small perturbations. Empirical results show thatour approach yields substantial improvements on a range of sentiment analysis,reasoning, and reading comprehension tasks. And CLINE also ensures thecompactness within the same semantics and separability across differentsemantics in sentence-level.

CLINE: Contrastive Learning with Semantic Negative Examples for Natural Language Understanding