Linguistics & Language · Computational Linguistics

Language Models as Linguistic Evidence

A topic within Computational Linguistics, itself one of 8 topics in that field and part of Linguistics & Language.

Reading on Language Models as Linguistic Evidence

2 works

Paper2016

Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies

Tal Linzen, Emmanuel Dupoux & Yoav Goldberg

Tests whether a recurrent network trained only on raw text learns long-distance subject-verb agreement, and uses the cases where it fails as evidence about what statistical learning alone can and cannot recover of syntax.

link checked 17 Sept 2026

Linguistics & Language · Computational Linguistics

FreeAdvanced

Paper2019

What Does BERT Look at? An Analysis of BERT's Attention

Clark, Khandelwal, Levy & Manning

Probes the attention heads inside a trained BERT model and finds several that track specific syntactic relations such as direct objects and coreference, without ever having been given a syntactic label during training.

Linguistics & Language · Computational Linguistics

FreeAdvanced

Reading on Language Models as Linguistic Evidence

Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies

What Does BERT Look at? An Analysis of BERT's Attention

Other topics in Computational Linguistics