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

Shauli Ravfogel, Grusha Prasad, Tal Linzen, Yoav Goldberg

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    When language models process syntactically complex sentences, do they use their representations of syntax in a manner that is consistent with the grammar of the language? We propose AlterRep, an intervention-based method to address this question. For any linguistic feature of a given sentence, AlterRep generates counterfactual representations by altering how the feature is encoded, while leaving intact all other aspects of the original representation. By measuring the change in a model’s word prediction behavior when these counterfactual representations are substituted for the original ones, we can draw conclusions about the causal effect of the linguistic feature in question on the model’s behavior. We apply this method to study how BERT models of different sizes process relative clauses (RCs). We find that BERT variants use RC boundary information during word prediction in a manner that is consistent with the rules of English grammar; this RC boundary information generalizes to a considerable extent across different RC types, suggesting that BERT represents RCs as an abstract linguistic category.

    Original languageEnglish (US)
    Title of host publicationCoNLL 2021 - 25th Conference on Computational Natural Language Learning, Proceedings
    EditorsArianna Bisazza, Omri Abend
    PublisherAssociation for Computational Linguistics (ACL)
    Pages194-209
    Number of pages16
    ISBN (Electronic)9781955917056
    StatePublished - 2021
    Event25th Conference on Computational Natural Language Learning, CoNLL 2021 - Virtual, Online
    Duration: Nov 10 2021Nov 11 2021

    Publication series

    NameCoNLL 2021 - 25th Conference on Computational Natural Language Learning, Proceedings

    Conference

    Conference25th Conference on Computational Natural Language Learning, CoNLL 2021
    CityVirtual, Online
    Period11/10/2111/11/21

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Human-Computer Interaction
    • Linguistics and Language

    Fingerprint

    Dive into the research topics of 'Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Prediction'. Together they form a unique fingerprint.

    Cite this