-Deep Learning has significantly improved the performance over time of the following long-standing components in speech processing systems

Encoder

Decoder

Speaker Embedding

Phoneme Representation

Feature Extraction

Hypothesis Search

Acoustic Model

Variability Compensation

Language Model

Backend Classification

-A traditional speech recognition system consists of the following components:

Feature Extraction

Encoder

Attention Mechanism

Decoder

Speaker Embedding

Acoustic Model

Language Model

Hypothesis Search