English

Robustifying Sentiment Classification by Maximally Exploiting Few Counterfactuals

Computation and Language 2022-10-24 v1

Abstract

For text classification tasks, finetuned language models perform remarkably well. Yet, they tend to rely on spurious patterns in training data, thus limiting their performance on out-of-distribution (OOD) test data. Among recent models aiming to avoid this spurious pattern problem, adding extra counterfactual samples to the training data has proven to be very effective. Yet, counterfactual data generation is costly since it relies on human annotation. Thus, we propose a novel solution that only requires annotation of a small fraction (e.g., 1%) of the original training data, and uses automatic generation of extra counterfactuals in an encoding vector space. We demonstrate the effectiveness of our approach in sentiment classification, using IMDb data for training and other sets for OOD tests (i.e., Amazon, SemEval and Yelp). We achieve noticeable accuracy improvements by adding only 1% manual counterfactuals: +3% compared to adding +100% in-distribution training samples, +1.3% compared to alternate counterfactual approaches.

Keywords

Cite

@article{arxiv.2210.11805,
  title  = {Robustifying Sentiment Classification by Maximally Exploiting Few Counterfactuals},
  author = {Maarten De Raedt and Fréderic Godin and Chris Develder and Thomas Demeester},
  journal= {arXiv preprint arXiv:2210.11805},
  year   = {2022}
}

Comments

EMNLP 2022

R2 v1 2026-06-28T04:09:38.071Z