CSE6328

This is an old revision of the document!

• Week 1: background introduction; speech and spoken language (Weekly Reading: W1)

• Week 2: Math foundation: probabilities; Bayes theorem; statistics; Entropy; mutual information; decision tree; optimization (Weekly Reading: W2) (A useful online manual on Matrix Calculus)

• Week 3: Bayesian decision theory; Discriminative Models: Linear discriminant functions, support vector machine (SVM); (Weekly Reading: W3_1(part I), W3_2(part II), W3_3(part III))

• Week 4: Generative Models; model estimation; maximum likelihood, EM algorithm; multivariate Gaussian, Gaussian mixture model, Multinomial, Markov Chain model; (Weekly Reading: W4)

• Week 5: Discriminative Learning; Bayesian Learning; Pattern Verification

• Week 6: Hidden Markov Model (HMM): HMM vs. Markov chains; HMM concepts; Three algorithms (forward-backward, Viterbi decoding, Baum-Welch learning) (Weekly Reading: W6, the full version of this paper is not required but is here for your reference)

• Week 8: Automatic Speech Recognition (ASR) (I): ASR introduction; ASR as an example of pattern classification; Acoustic modeling: parameter tying (decision tree based state tying); (Weekly Reading: W8)

• Week 9: Automatic Speech Recognition (ASR) (II): Language Modelling (LM); N-gram models: smoothing, learning, perplexity, class-based.

• Week 10: Automatic Speech Recognition (ASR) (III): Search - why search; Search space in n-gram LM; Viterbi decoding in a large HMM; beam search; tree-based lexicon; dynamic decoding; static decoding; weighted finite state transducer (WFST) (Additional slides for WFST)