CS E6998-1: Speech Processing

CS E6998-1: Speech Processing
Course Outline

    1. Introduction to speech recognition and introduction to the theory of automata
      • Course topic
      • Course information
      • The problem of speech recognition
      • Introduction to automata
    2. Automata-theoretic foundations
      • Regular or rational set, finite automata
      • Rational relations, rational transductions
    3. Introduction to the finite-state software tools
      • Description of the FSM Library
      • Formats and representations
      • Utilities
    4. Rational relations, transductions, and weighted finite-state transducers
      • Rational relations, rational transductions
      • Weighted automata and transducers
    5. Elementary weighted automata algorithms
      • Rational operations (concatenation, union, closure)
      • Other operations (reversal, complementation, projection, inversion)
      • Lazy implementations
    6. Advanced weighted automata algorithms (1)
      • Elementary graph algorithms
      • Composition (weighted acceptors and transducers)
      • Intersection (weighted acceptors)
      • Difference (weighted and unweighted acceptors)
    7. Advanced weighted automata algorithms (2)
      • Determinization (weighted acceptors and transducers)
      • Minimization (weighted acceptors and transducers)
    8. Advanced weighted automata algorithms (3)
      • Minimization (weighted acceptors and transducers)
      • Equality (weighted acceptors and transducers)
      • Equivalence (weighted acceptors and transducers)
    9. Shortest-paths algorithms
      • Single-source shortest paths algorithms
      • Single-source shortest distance algorithms
      • All-pairs shortest paths algorithms
      • Epsilon-removal (weighted acceptors and transducers)
    10. Speech recognition by composition of weighted transducers
      • Components of a speech recognition system
      • Combination and search
    11. Models in speech recognition
      • Language models
      • Pronunciation models
      • Context-dependent phone models
    12. Large vocabulary speech recognition
      • Optimizations
      • Dynamic modifications
      • Word and phone lattices
      • Current issues and results
    13. Text-to-speech
      • Weighted transducers in TTS systems
      • Weighted context-dependent rules
      • Local grammars
    14. Unifying framework
      • Semirings
      • Rational power series
      • Characterizations
      • General properties