Perturbation-based Active Learning for Question Answering
This work addresses the challenge of efficient data annotation for question answering systems, but it appears incremental as it builds on existing active learning methods.
The authors tackled the problem of reducing annotation costs in question answering models by proposing a perturbation-based active learning acquisition strategy, which they demonstrated to be more effective than existing common strategies.
Building a question answering (QA) model with less annotation costs can be achieved by utilizing active learning (AL) training strategy. It selects the most informative unlabeled training data to update the model effectively. Acquisition functions for AL are used to determine how informative each training example is, such as uncertainty or diversity based sampling. In this work, we propose a perturbation-based active learning acquisition strategy and demonstrate it is more effective than existing commonly used strategies.