In the past few years, significant progress has been made in both research and commercial applications of speech and language processing. However, there remain important theoretical issues to be addressed on statistical modeling and learning. Theoretical advancement is expected to drive greater system performance improvement, which in turn generates the new need of novel learning and modeling methodologies as well as in-depth studies of them.