New Datasets to Democratize Speech Recognition Technology