# Selection Bias Explorations and Debias Methods for Natural Language Sentence Matching Datasets

arXiv: https://arxiv.org/abs/1905.06221 · HTML: https://mlsanity.com/paper/1905.06221/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 May 2019.

## Built on by (most cited first, 8 of 8)

- [Uncertainty Calibration for Ensemble-Based Debiasing Methods](https://mlsanity.com/paper/2111.04104/) (2021)
- [Blow the Dog Whistle: A Chinese Dataset for Cant Understanding with Common Sense and World Knowledge](https://mlsanity.com/paper/2104.02704/) (2021)
- [PECO: Examining Single Sentence Label Leakage in Natural Language Inference Datasets through Progressive Evaluation of Cluster Outliers](https://mlsanity.com/paper/2112.09237/) (2021)
- [Explanatory Pluralism in Explainable AI](https://mlsanity.com/paper/2106.13976/) (2021)
- [Less is Better: Recovering Intended-Feature Subspace to Robustify NLU Models](https://mlsanity.com/paper/2209.07879/) (2022)
- [Biases in Scholarly Recommender Systems: Impact, Prevalence, and Mitigation](https://mlsanity.com/paper/2301.07483/) (2023)
- [DCAST: Diverse Class-Aware Self-Training Mitigates Selection Bias for Fairer Learning](https://mlsanity.com/paper/2409.20126/) (2024)
- [Multi-Objective Few-shot Learning for Fair Classification](https://mlsanity.com/paper/2110.01951/) (2021)

## Abstract (arXiv)

Natural Language Sentence Matching (NLSM) has gained substantial attention from both academics and the industry, and rich public datasets contribute a lot to this process. However, biased datasets can also hurt the generalization performance of trained models and give untrustworthy evaluation results. For many NLSM datasets, the providers select some pairs of sentences into the datasets, and this sampling procedure can easily bring unintended pattern, i.e., selection bias. One example is the QuoraQP dataset, where some content-independent naive features are unreasonably predictive. Such features are the reflection of the selection bias and termed as the leakage features. In this paper, we investigate the problem of selection bias on six NLSM datasets and find that four out of them are significantly biased. We further propose a training and evaluation framework to alleviate the bias. Experimental results on QuoraQP suggest that the proposed framework can improve the generalization ability of trained models, and give more trustworthy evaluation results for real-world adoptions.