A Toolkit for Detecting Spurious Correlations in Speech Datasets
For researchers developing speech-based health systems, this toolkit helps identify dangerous overestimations of model performance due to spurious correlations.
The paper introduces a toolkit to detect spurious correlations between recording conditions and target class in speech datasets, using non-speech regions to flag performance overestimation. The toolkit is publicly available.
We introduce a toolkit for uncovering spurious correlations between recording characteristics and target class in speech datasets. Spurious correlations may arise due to heterogeneous recording conditions, a common scenario for health-related datasets. When present both in the training and test data, these correlations result in an overestimation of the system performance -- a dangerous situation, specially in high-stakes application where systems are required to satisfy minimum performance requirements. Our toolkit implements a diagnostic method based on the detection of the target class using only the non-speech regions in the audio. Better than chance performance at this task indicates that information about the target class can be extracted from the non-speech regions, flagging the presence of spurious correlations. The toolkit is publicly available for research use.