Naturalistic Causal Probing for Morpho-Syntax

TL;DR

This paper introduces a naturalistic causal probing method that intervenes on morpho-syntactic features in sentences to causally analyze pre-trained language models, revealing stable effects of linguistic properties.

Contribution

It proposes a novel input-level intervention strategy for causal probing of models using naturalistic sentences, enhancing understanding of linguistic feature effects.

Findings

Interventions yield stable estimates of linguistic effects

Grammatical gender and number influence model representations

Naturalistic probing improves interpretability of pre-trained models

Abstract

Probing has become a go-to methodology for interpreting and analyzing deep neural models in natural language processing. However, there is still a lack of understanding of the limitations and weaknesses of various types of probes. In this work, we suggest a strategy for input-level intervention on naturalistic sentences. Using our approach, we intervene on the morpho-syntactic features of a sentence, while keeping the rest of the sentence unchanged. Such an intervention allows us to causally probe pre-trained models. We apply our naturalistic causal probing framework to analyze the effects of grammatical gender and number on contextualized representations extracted from three pre-trained models in Spanish: the multilingual versions of BERT, RoBERTa, and GPT-2. Our experiments suggest that naturalistic interventions lead to stable estimates of the causal effects of various linguistic properties. Moreover, our experiments demonstrate the importance of naturalistic causal probing when analyzing pre-trained models.