Article
G2P Shrinks Speech Models
By
ā¢
ā¢
23Interesting, but how does this approach generalize to arbitrary user query / document domains? Would you need to train a separate network for each domain / dataset?