Disfl-QA: A Benchmark Dataset for Understanding Disfluencies in Question Answering
This addresses the lack of datasets for disfluencies in NLP, which is crucial for making QA models robust to real-world conversational noise, though it is incremental as it builds on SQuAD.
The authors tackled the problem of disfluencies in NLP by creating Disfl-QA, a benchmark dataset derived from SQuAD with human-introduced disfluencies, and found that state-of-the-art QA models degrade significantly in zero-shot testing, with data augmentation partially recovering performance.
Disfluencies is an under-studied topic in NLP, even though it is ubiquitous in human conversation. This is largely due to the lack of datasets containing disfluencies. In this paper, we present a new challenge question answering dataset, Disfl-QA, a derivative of SQuAD, where humans introduce contextual disfluencies in previously fluent questions. Disfl-QA contains a variety of challenging disfluencies that require a more comprehensive understanding of the text than what was necessary in prior datasets. Experiments show that the performance of existing state-of-the-art question answering models degrades significantly when tested on Disfl-QA in a zero-shot setting.We show data augmentation methods partially recover the loss in performance and also demonstrate the efficacy of using gold data for fine-tuning. We argue that we need large-scale disfluency datasets in order for NLP models to be robust to them. The dataset is publicly available at: https://github.com/google-research-datasets/disfl-qa.