Mask-Net: Learning Context Aware Invariant Features Using Adversarial Forgetting (Student Abstract)

Training a robust system, e.g., Speech to Text (STT), requires large datasets. Variability present in the dataset, such as unwanted nuances and biases, is reason for need datasets learn general representations. In this work, we propose novel approach induce invariance using adversarial forgetting (AF). Our initial experiments on learning invariant features accent STT task achieve better generalizations terms of word error rate (WER) compared traditional models. We observe an absolute improvement 2.2% 1.3% out-of-distribution in-distribution test sets, respectively.