A good friend of mine has been working on low resources NLP for a couple years now. A few months ago, he joined Mila Quebec AI Institute under Yoshua Bengio. Where it is hard to find transcribed audio, regular audio is just as hard to come by... Until you start mining all those forwarded audios on WhatsApp.
I used to get mad every time my mom would forward a 30 minutes audio that she knows I would delete before opening. And I used to delete them all. Now, I download the audio. Label the language it is in, and add it to my friends shared training data folder.
I used to get mad every time my mom would forward a 30 minutes audio that she knows I would delete before opening. And I used to delete them all. Now, I download the audio. Label the language it is in, and add it to my friends shared training data folder.