Privacy-preserving Active Learning on Sensitive Data for User Intent Classification [article]

Oluwaseyi Feyisetan, Thomas Drake, Borja Balle, Tom Diethe
2019 arXiv   pre-print
Active learning holds promise of significantly reducing data annotation costs while maintaining reasonable model performance. However, it requires sending data to annotators for labeling. This presents a possible privacy leak when the training set includes sensitive user data. In this paper, we describe an approach for carrying out privacy preserving active learning with quantifiable guarantees. We evaluate our approach by showing the tradeoff between privacy, utility and annotation budget on a
more » ... binary classification task in a active learning setting.
arXiv:1903.11112v1 fatcat:a6fssjx4a5aq5giif4g44qh4mu