Decoding Probing: Revealing Internal Linguistic Structures in Neural Language Models using Minimal Pairs

TL;DR

This paper introduces a decoding probing method using minimal pairs to analyze internal linguistic structures in neural language models, revealing how different models and layers encode grammatical and semantic information.

Contribution

The study presents a novel decoding probing approach with minimal pairs to uncover layer-wise linguistic representations in neural language models, highlighting differences from traditional models.

Findings

Self-supervised models capture abstract linguistic structures in intermediate layers.

Grammaticality information is robustly encoded in early layers of GPT-2.

Morphological and semantics-syntax interface features are harder to decode.

Abstract

Inspired by cognitive neuroscience studies, we introduce a novel `decoding probing' method that uses minimal pairs benchmark (BLiMP) to probe internal linguistic characteristics in neural language models layer by layer. By treating the language model as the `brain' and its representations as `neural activations', we decode grammaticality labels of minimal pairs from the intermediate layers' representations. This approach reveals: 1) Self-supervised language models capture abstract linguistic structures in intermediate layers that GloVe and RNN language models cannot learn. 2) Information about syntactic grammaticality is robustly captured through the first third layers of GPT-2 and also distributed in later layers. As sentence complexity increases, more layers are required for learning grammatical capabilities. 3) Morphological and semantics/syntax interface-related features are harder to capture than syntax. 4) For Transformer-based models, both embeddings and attentions capture grammatical features but show distinct patterns. Different attention heads exhibit similar tendencies toward various linguistic phenomena, but with varied contributions.