Fast on-the-fly composition for weighted finite-state transducers in 1.8 million-word vocabulary continuous speech recognition.

Takaaki Hori, Chiori Hori, Yasuhiro Minami

Browse the full Interspeech paper archive.