Mathematical Models for Speech Technology

Author's preface.

1 Introduction

2 Preliminaries

2.1 The physics of speech production

2.2 The source-filter model

2.3 Information-bearing features of the speech signal

2.4 Time-frequency representations

2.5 Classifications of acoustic patterns in speech

2.6 Temporal invariance and stationarity

2.7 Taxonomy of linguistic structure

3 Mathematical models of linguistic structure

3.1 Probabilistic functions of a discrete Markov process

3.2 Formal grammars and abstract automata

4 Syntactic analysis

4.1 Deterministic parsing algorithms

4.2 Probabilistic parsing algorithms

4.3 Parsing natural language

5 Grammatical inference

5.1 Exact inference and Gold's theorem

5.2 Baum's algorithm for regular grammars

5.3 Event counting in parse trees

5.4 Baker's algorithm for context-free grammars

6 Information-theoretic analysis of speech communication

6.1 The Miller et al. experiments

6.2 Entropy of an information source

6.3 Recognition error rates and entropy

7 Automatic speech recognition and constructive theories of language

7.1 Integrated architectures

7.2 Modular architectures

7.3 Parameter estimation from fluent speech

7.4 System performance

7.5 Other speech technologies

8 Automatic speech understanding and semantics

8.1 Transcription and comprehension

8.2 Limited domain semantics

8.3 The semantics of natural language

8.4 System architectures

8.5 Human and machine performance

9 Theories of mind and language

9.1 The challenge of automatic natural language understanding

9.2 Metaphors for mind

9.3 The artificial intelligence program

10 A speculation on the prospects for a science of the mind

10.1 The parable of the thermos bottle: measurements and symbols

10.2 The four questions of science

10.3 A constructive theory of the mind

10.4 The problem of consciousness

10.5 The role of sensorimotor function, associative memory and reinforcement learning in automatic acquisition of spoken language by an autonomous robot

10.6 Final thoughts: predicting the course of discovery