speech-to-intent-dataset - 1.0

This release covers 14 coarse-grained intents from the Banking domain. The dataset is split into:

• test - 100 samples per intent
• train - >650 samples per intent

Audios were generated by 11 (Indian English) speakers, recording over a telephony line. Also includes anonymised speaker information - like native language, languages spoken, native language and gender - to allow more structured discussions around robustness and bias.

- Python
Published by janaab11 over 3 years ago