What are the key differences between CTC-based, Transducer-based (RNN-T), and Attention-based Encoder-Decoder (AED) models for speech recognition?
