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

When language models process syntactically complex sentences, do they usetheir representations of syntax in a manner that is consistent with the grammarof the language? We propose AlterRep, an intervention-based method to addressthis question. For any linguistic feature of a given sentence, AlterRepgenerates counterfactual representations by altering how the feature isencoded, while leaving intact all other aspects of the original representation.By measuring the change in a model's word prediction behavior when thesecounterfactual representations are substituted for the original ones, we candraw conclusions about the causal effect of the linguistic feature in questionon the model's behavior. We apply this method to study how BERT models ofdifferent sizes process relative clauses (RCs). We find that BERT variants useRC boundary information during word prediction in a manner that is consistentwith the rules of English grammar; this RC boundary information generalizes toa considerable extent across different RC types, suggesting that BERTrepresents RCs as an abstract linguistic category.

Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Prediction