References & Citations
Computer Science > Computation and Language
Title: Naturalistic Causal Probing for Morpho-Syntax
(Submitted on 14 May 2022 (this version), latest version 14 Nov 2022 (v2))
Abstract: Probing has become a go-to methodology for interpreting and analyzing deep neural models in natural language processing. Yet recently, there has been much debate around the limitations and weaknesses of probes. In this work, we suggest a naturalistic strategy for input-level intervention on real world data in Spanish, which is a language with gender marking. Using our approach, we isolate morpho-syntactic features from counfounders in sentences, e.g. topic, which will then allow us to causally probe pre-trained models. We apply this methodology to analyze causal effects of gender and number on contextualized representations extracted from pre-trained models -- BERT, RoBERTa and GPT-2. Our experiments suggest that naturalistic intervention can give us stable estimates of causal effects, which varies across different words in a sentence. We further show the utility of our estimator in investigating gender bias in adjectives, and answering counterfactual questions in masked prediction. Our probing experiments highlights the importance of conducting causal probing in determining if a particular property is encoded in representations.
Submission history
From: Afra Amini [view email][v1] Sat, 14 May 2022 11:47:58 GMT (3092kb,D)
[v2] Mon, 14 Nov 2022 11:41:52 GMT (1301kb,D)
Link back to: arXiv, form interface, contact.