Google has published its Speech Commands Dataset consisting of 65,000 one-second long clips of thousands of people speaking 30 short words, such as “yes,” “on,” and “stop.” Google developed the Speech Commands Dataset to serve as training data for speech recognition albums powering basic voice interfaces. Google also made the application it used to create this dataset available as open source to encourage developers to create similar systems in other languages and with bigger vocabularies.
Image: OpenClipart-Vectors.