Retrieval-guided Counterfactual Generation for QA

TL;DR

This paper introduces RGF, a retrieval-based method for generating diverse, fluent counterfactual questions to improve QA model robustness and performance, especially on out-of-domain and challenging datasets.

Contribution

We develop a Retrieve-Generate-Filter technique for automatic counterfactual data creation in QA, reducing human effort and enhancing model robustness.

Findings

RGF-generated data improves QA performance on out-of-domain datasets

Counterfactual augmentation enhances model robustness to local perturbations

Our method outperforms existing approaches in generating diverse, fluent counterfactuals

Abstract

Deep NLP models have been shown to learn spurious correlations, leaving them brittle to input perturbations. Recent work has shown that counterfactual or contrastive data -- i.e. minimally perturbed inputs -- can reveal these weaknesses, and that data augmentation using counterfactuals can help ameliorate them. Proposed techniques for generating counterfactuals rely on human annotations, perturbations based on simple heuristics, and meaning representation frameworks. We focus on the task of creating counterfactuals for question answering, which presents unique challenges related to world knowledge, semantic diversity, and answerability. To address these challenges, we develop a Retrieve-Generate-Filter(RGF) technique to create counterfactual evaluation and training data with minimal human supervision. Using an open-domain QA framework and question generation model trained on original task data, we create counterfactuals that are fluent, semantically diverse, and automatically labeled. Data augmentation with RGF counterfactuals improves performance on out-of-domain and challenging evaluation sets over and above existing methods, in both the reading comprehension and open-domain QA settings. Moreover, we find that RGF data leads to significant improvements in a model's robustness to local perturbations.