Skip to main navigation Skip to search Skip to main content

How is BERT surprised? Layerwise detection of linguistic anomalies

  • Bai Li
  • , Zining Zhu
  • , Guillaume Thomas
  • , Yang Xu
  • , Frank Rudzicz
    • University of Toronto

    Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

    21 Scopus citations

    Abstract

    Transformer language models have shown remarkable ability in detecting when a word is anomalous in context, but likelihood scores offer no information about the cause of the anomaly. In this work, we use Gaussian models for density estimation at intermediate layers of three language models (BERT, RoBERTa, and XLNet), and evaluate our method on BLiMP, a grammaticality judgement benchmark. In lower layers, surprisal is highly correlated to low token frequency, but this correlation diminishes in upper layers. Next, we gather datasets of morphosyntactic, semantic, and commonsense anomalies from psycholinguistic studies; we find that the best performing model RoBERTa exhibits surprisal in earlier layers when the anomaly is morphosyntactic than when it is semantic, while commonsense anomalies do not exhibit surprisal at any intermediate layer. These results suggest that language models employ separate mechanisms to detect different types of linguistic anomalies.

    Original languageEnglish
    Title of host publicationACL-IJCNLP 2021 - 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, Proceedings of the Conference
    Pages4215-4228
    Number of pages14
    ISBN (Electronic)9781954085527
    DOIs
    StatePublished - 2021
    EventJoint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL-IJCNLP 2021 - Virtual, Online
    Duration: 1 Aug 2021 → 6 Aug 2021

    Publication series

    NameACL-IJCNLP 2021 - 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, Proceedings of the Conference
    Volume1

    Conference

    ConferenceJoint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL-IJCNLP 2021
    CityVirtual, Online
    Period1/08/21 → 6/08/21

    Fingerprint

    Dive into the research topics of 'How is BERT surprised? Layerwise detection of linguistic anomalies'. Together they form a unique fingerprint.

    Cite this