ECE 590SIP, Fall 2020

Image Source

Deep F-measure Maximization for End-to-End Speech Understanding

Leda Sari, September 21, 2020, 4:00-4:30pm CDT

Spoken language understanding (SLU) datasets, like manyother machine learning datasets, usually suffer from the label imbalance problem. Label imbalance usually causes the learned model to replicate similar biases at the output which raises the issue of unfairness to the minority classes in the dataset. In thiswork, we approach the fairness problem by maximizing the F-measure instead of accuracy in neural network model training. We propose a differentiable approximation to the F-measure and train the network with this objective using standard back-propagation. We perform experiments on two standard fairness datasets, Adult, and Communities and Crime, and also on speech-to-intent detection on the ATIS dataset and speech-to-image concept classification on the Speech-COCO dataset. In all four of these tasks, F-measure maximization results in improved micro-F1 scores, with absolute improvements of up to 8% absolute, as compared to models trained with the cross-entropy loss function. In the two multi-class SLU tasks, theproposed approach significantly improves class coverage, i.e., the number of classes with positive recall.

Link to the paper