
Acoustic Word Embeddings for Low Resource Speech Processing with Herman Kamper - TWiML Talk #191
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Is There an Iterative Cycle?
Rit. Oyushoulde, you should have written the abstract for my thesis. So what you just said is exactly right. You start with like a random sigmentation of your impud corpus and then on that random Sigmentation, you buld an embedding space. If i haven't an imbedding space, then what you can do is you can say, given this embedding space, how should i split up my imped stream to cind of havea higher score? Cause, so you go back, you resigment, it's elimit. Lmet, let me pas you there. Exactly. I us gashyour mixture modelty wi, the clust
Transcript
Play full episode