3 ms·

Why does data for speech recognition need to be prefect. That's certainly not the case for other machine learning applications. Can you train the less clean dat

Why does data for speech recognition need to be prefect. That's certainly not the case for other machine learning applications. Can you train the less clean data and fine-tune on a clean subset?

dabinat 5y ago
Well that was kind of my point: you need to manually figure out what’s clean and what isn’t.
- ma2rten 5y ago
  But it's easy to do that for a small subset for finetuning compared to cleaning up the entire dataset.