• Typical LVCSR systems have about 10M free parameters, which makes training a challenge. • Large  speech databases are required (several hundred hours of speech). • Tying, smoothing, and interpolation are required.