CMSC 35900
Topics in Artificial Intelligence: Speech Technologies

Autumn 2009

Course schedule: TTh 1:30-2:50 at TTI Room #526 (6045 S. Kenwood Ave)

Course textbook: None (selected readings will be provided)

Instructor: Karen Livescu

Prerequisites: A good understanding of probability

Course description:

This course will introduce techniques used in speech technologies, mainly focusing on speech recognition. Speech recognition is one of the oldest and most complex structured sequence prediction tasks receiving significant research and commercial attention, and therefore provides a good case study for many of the techniques that are used in other areas of artificial intelligence involving sequence prediction. Current work in speech technologies is also a good example of the effectiveness of combining statistics and learning with domain knowledge (from linguistics and acoustics). Sample topics to be covered: Feature extraction, phonetic classification, acoustic modeling with hidden Markov models, pronunciation modeling, language modeling, expectation-maximization for learning, large-vocabulary recognition, discriminative models, graphical models, adaptation to speaker/acoustic condition.

For students