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

While vector-based language representations from pretrained language modelshave set a new standard for many NLP tasks, there is not yet a completeaccounting of their inner workings. In particular, it is not entirely clearwhat aspects of sentence-level syntax are captured by these representations,nor how (if at all) they are built along the stacked layers of the network. Inthis paper, we aim to address such questions with a general class ofinterventional, input perturbation-based analyses of representations frompretrained language models. Importing from computational and cognitiveneuroscience the notion of representational invariance, we perform a series ofprobes designed to test the sensitivity of these representations to severalkinds of structure in sentences. Each probe involves swapping words in asentence and comparing the representations from perturbed sentences against theoriginal. We experiment with three different perturbations: (1) randompermutations of n-grams of varying width, to test the scale at which arepresentation is sensitive to word position; (2) swapping of two spans whichdo or do not form a syntactic phrase, to test sensitivity to global phrasestructure; and (3) swapping of two adjacent words which do or do not breakapart a syntactic phrase, to test sensitivity to local phrase structure.  Results from these probes collectively suggest that Transformers buildsensitivity to larger parts of the sentence along their layers, and thathierarchical phrase structure plays a role in this process. More broadly, ourresults also indicate that structured input perturbations widens the scope ofanalyses that can be performed on often-opaque deep learning systems, and canserve as a complement to existing tools (such as supervised linear probes) forinterpreting complex black-box models.

Syntactic Perturbations Reveal Representational Correlates of Hierarchical Phrase Structure in Pretrained Language Models