Selection Bias Explorations and Debias Methods for Natural Language Sentence Matching Datasets
This addresses the problem of biased datasets hurting generalization and evaluation reliability in natural language sentence matching for researchers and practitioners, though it is incremental as it builds on existing debias methods.
The paper investigates selection bias in six natural language sentence matching datasets, finding significant bias in four of them, and proposes a training and evaluation framework that improves model generalization and provides more trustworthy evaluation results, as demonstrated on the QuoraQP dataset.
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.