|
Chapter 14: Learning a production grammar
|
|
Learners start with empty grammars, and have to learn both the constraints
and their rankings. A convergent and robust gradual learning algorithm exists.
Contents
14.1 Grammar model
14.2 Learning in functional phonology
14.2.1 The grammar
14.2.2 Gradual learning algorithms
14.2.3 Three production modes
14.2.4 Stage 1: an empty grammar
14.2.5 Step 1: perceptual categorization and faithfulness constraints
14.2.6 Stage 2: violated faithfulness
14.2.7 Step 2: sensorimotor learning
14.2.8 Stage 3: faithful imitation
14.2.9 Step 3: the learning curve
14.2.10 Stage 4: faithfulness outranks gestural constraints
14.2.11 Step 4: sentence-level phonology
14.2.12 Stage 5: alternating levels of constraints
14.2.13 Step 5: emergence of underlying forms
14.2.14 Stage 6: the adult phase
14.2.15 Second-language acquisition
14.2.16 Acoustic versus linguistic faithfulness
14.2.17 Puzzles
14.3 Example: acquisition of tongue-root harmony
14.3.1 Universal ranking of articulatory constraints
14.3.2 Universal ranking of faithfulness constraints
14.3.3 Typology of tongue-root systems
14.3.4 The learning process for continuous families
14.3.5 The learning of simplified Wolof
14.3.6 An alternative Wolof: articulatory versus perceptual candidates
14.3.7 Wolof with schwa licensing
14.3.8 Learning unnatural local rankings
14.3.9 Real Wolof
14.4 Principles-and-parameters learning algorithms
14.4.1 Seven possible tongue-root-harmony systems
14.4.2 The Triggering Learning Algorithm
14.4.3 The subset problem
14.4.4 Intermezzo: the correct P&P convergence criterion
14.4.5 Local maxima
14.4.6 Relaxing conservatism or greediness
14.4.7 Genetic algorithms
14.4.8 TLA versus GLA
14.5 Optimality-theoretic learning
14.5.1 The initial state in an Optimality-theoretic grammar
14.5.2 Innateness
14.6 Algorithm
14.7 Proof of learnability
14.8 Acquisition time
14.9 Conclusion
Forward to chapter 15.
Up to the Functional Phonology web site.