| Foreword | p. xi |
| Introduction | p. 1 |
| What this book is about | p. 1 |
| Speech recognition and language models | p. 5 |
| What Regulus does | p. 13 |
| Clarissa and MedSLT | p. 15 |
| Related work | p. 20 |
| Plan of the book | p. 20 |
| Summary | p. 21 |
| Using Regulus | p. 23 |
| Getting started | p. 25 |
| Getting set up | p. 25 |
| A toy grammar in GSL | p. 28 |
| Rewriting Toy0 in Regulus | p. 32 |
| Regulus configuration files | p. 37 |
| Using Regulus | p. 39 |
| Summary | p. 40 |
| Simple applications | p. 43 |
| Introduction | p. 43 |
| The Regulus Speech Server | p. 44 |
| A toy dialogue system in Prolog | p. 46 |
| A toy speech translation system in Prolog | p. 50 |
| A toy dialogue system in Java | p. 53 |
| Summary | p. 62 |
| Developing grammars | p. 65 |
| Introduction | p. 65 |
| Using the Regulus development environment | p. 65 |
| The Toy1 example grammar | p. 67 |
| Unification | p. 77 |
| Macros | p. 81 |
| Compiling the Toy1 recogniser | p. 85 |
| Systematic testing of recognisers | p. 87 |
| Summary | p. 89 |
| A spoken dialogue system | p. 93 |
| Introduction | p. 93 |
| The Toy1 spoken dialogue system | p. 95 |
| The input manager | p. 102 |
| The dialogue manager | p. 104 |
| The output manager | p. 108 |
| Integrating dialogue management with recognition | p. 108 |
| Dealing with ellipsis and corrections | p. 112 |
| Summary | p. 117 |
| A speech translation system | p. 119 |
| Introduction | p. 119 |
| Transfer-based systems | p. 120 |
| Developing translation applications | p. 127 |
| Translation through interlingua | p. 132 |
| Translation of ellipsis | p. 134 |
| Systematic development | p. 138 |
| Integrating translation with recognition | p. 142 |
| Summary | p. 145 |
| Using grammar specialisation | p. 149 |
| Overview | p. 149 |
| Using the general English grammar | p. 150 |
| The training corpus | p. 154 |
| Adding lexical entries | p. 156 |
| General grammar semantics | p. 166 |
| Multiple top-level specialised grammars | p. 169 |
| Including lexicon entries directly | p. 169 |
| Dealing with ambiguity | p. 171 |
| Making compilation more efficient | p. 171 |
| Using probabilistic tuning | p. 172 |
| Summary | p. 173 |
| How Regulus Works | p. 175 |
| Compiling feature grammars into CFG | p. 177 |
| Introduction | p. 177 |
| Exhaustive expansion | p. 178 |
| Filtering | p. 179 |
| Efficient filtering of CFGs | p. 182 |
| Interleaving expansion and filtering | p. 186 |
| Pre-processing of feature grammars | p. 195 |
| Transforming the output CFG | p. 199 |
| Semantics | p. 203 |
| Summary | p. 203 |
| A general English feature grammar for speech | p. 205 |
| Introduction | p. 205 |
| What makes speech grammars special | p. 206 |
| English grammar: basic intuitions | p. 206 |
| Compositional semantics | p. 209 |
| Noun phrases | p. 211 |
| Verb phrases and basic clauses | p. 214 |
| Adjuncts | p. 228 |
| Coordination | p. 229 |
| Feature defaults | p. 230 |
| Summary | p. 231 |
| Grammar specialisation using Explanation Based Learning | p. 233 |
| Explanation Based Learning | p. 233 |
| Defining cutting-up criteria | p. 244 |
| Different kinds of cutting-up criteria | p. 246 |
| Summary | p. 251 |
| Performance of grammar-based recognisers | p. 255 |
| Introduction | p. 255 |
| Varying vocabulary size | p. 256 |
| Varying linguistic coverage | p. 259 |
| Varying the feature set | p. 261 |
| Varying the cutting-up criteria | p. 263 |
| Comparing CFG and PCFG language models | p. 266 |
| Deriving recognisers from general grammars | p. 267 |
| Summary | p. 268 |
| Comparison of rule-based and robust approaches | p. 271 |
| Introduction | p. 271 |
| Methodological issues | p. 272 |
| Experiments on MedSLT | p. 279 |
| Experiments on Clarissa | p. 281 |
| Discussion | p. 282 |
| Summary | p. 286 |
| Summary and future directions | p. 289 |
| Summary | p. 289 |
| Future directions | p. 291 |
| Online Documentation | p. 293 |
| References | p. 295 |
| Index | p. 301 |
| Table of Contents provided by Ingram. All Rights Reserved. |