Encoder-Decoder ASR
An offline speech recognition model that reads the entire audio clip to grasp the full context before translating it to text word by word using cross-attention.
Think of It Like This
Like a translator listening to a full recorded speech before writing down the translation, allowing them to use future context to decipher mumbled words.
Encoder-decoder architectures use two distinct components: an encoder that compresses the full audio into context-rich representations, and a decoder that writes out the transcript word by word. They provide high accuracy because they look at the whole sequence, but because they must wait for the audio to finish, they cannot stream live.