Table of Contents
Acknowledgments xi
Setting the Stage xiii
1 Our Vision of Linguistics for the Age of AI 1
1.1 What Is Linguistics for the Age of AI? 1
1.2 What Is So Hard about Language? 2
1.3 Relevant Aspects of the History of Natural Language Processing 4
1.4 The Four Pillars of Linguistics for the Age of AI 8
1.4.1 Pillar 1: Language Processing Capabilities Are Developed within an integrated, Comprehensive Agent Architecture 9
1.4.2 Pillar 2: Modeling Is Human-Inspired in Service of Explanatory AI and Actionability 13
1.4.3 Pillar 3: Insights Are Cleaned from Linguistic Scholarship and, in Turn, Contribute to That Scholarship 15
1.4.3.1 Theoretical syntax 16
1.4.3.2 Psycholinguistics 17
1.4.3.3 Semantics 17
1.4.3.4 Pragmatics 20
1.4.3.5 Cognitive linguistics 22
1.4.3.6 Language evolution 23
1.4.4 Pillar 4: All Available Heuristic Evidence Is Incorporated When Extracting and Representing the Meaning of Language Inputs 25
1.4.4.1 Handcrafted knowledge bases for NLP 25
1.4.4.2 Using results from empirical NLP 27
1.5 The Goals of This Book 29
1.6 Deep Dives 31
1.6.1 The Phenomenological Stance 31
1.6.2 Learning 33
1.6.3 NLP and NLU: It's Not Either-Or 33
1.6.4 Cognitive Systems: A Bird's-Eye View 36
1.6.5 Explanation in AI 38
1.6.6 Incrementality in the History of NLP 40
1.6.7 Why Machine-Readable, Human-Oriented Resources Are Not Enough 42
1.6.8 Coreference in the Knowledge-Lean Paradigm 44
1.6.9 Dialog Act Detection 46
1.6.10 Grounding 48
1.6.11 More on Empirical NLP 50
1.6.12 Manual Corpus Annotation: Its Contributions, Complexities, and Limitations 52
1.7 Further Exploration 55
2 A Brief Overview of Natural Language Understanding by LEIAs 59
2.1 Theory, Methodology, and Strategy 59
2.2 A Warm-Up Example 64
2.3 Knowledge Bases 68
2.3.1 The Ontology 69
2.3.2 The Lexicon 73
2.3.3 Episodic Memory 77
2.4 Incrementality 77
2.5 The Stages of NLU and Associated Decision-Making 79
2.5.1 Decision-Making after Pre-Semantic Analysis 84
2.5.2 Decision-Making after Pre-Semantic Integration 85
2.5.3 Decision-Making after Basic Semantic Analysis 85
2.5.4 Decision-Making after Basic Coreference Resolution 86
2.5.5 Decision-Making after Extended Semantic Analysis 87
2.5.6 Decision-Making after Situational Reasoning 87
2.6 Microtheories 88
2.7 "Golden" Text Meaning Representations 93
2.8 Deep Dives 94
2.8.1 The LEIA Knowledge Representation Language versus Other Options 94
2.8.2 Issues of Ontology 97
2.8.3 Issues of Lexicon 102
2.8.4 Paraphrase in Natural Language and the Ontological Metalanguage 111
2.9 Further Exploration 115
3 Pre-Semantic Analysis and Integration 117
3.1 Pre-Semantic Analysis 117
3.2 Pre-Semantic Integration 118
3.2.1 Syntactic Mapping: Basic Strategy 118
3.2.2 Recovering from Production Errors 124
3.2.3 Learning New Words and Word Senses 124
3.2.4 Optimizing Imperfect Syn-Maps 126
3.2.5 Reambiguating Certain Syntactic Decisions 128
3.2.6 Handling Known Types of Parsing Errors 129
3.2.7 From Recovery Algorithm to Engineering Strategy 131
3.3 Managing Combinatorial Complexity 134
3.4 Taking Stock 138
3.5 Further Exploration 139
4 Basic Semantic Analysis 141
4.1 Modification 143
4.1.1 Recorded Property Values 143
4.1.2 Dynamically Computed Values for Scalar Attributes 145
4.1.3 Modifiers Explained Using Combinations of Concepts 149
4.1.4 Dynamically Computed Values for Relative Text Components 150
4.1.5 Quantification and Sets 152
4.1.6 Indirect Modification 158
4.1.7 Recap of Modification 159
4.2 Proposition-Level Semantic Enhancements 160
4.2.1 Modality 160
4.2.2 Aspect 162
4.2.3 Non-Modal, Non-Aspectual Matrix Verbs 163
4.2.4 Questions 163
4.2.5 Commands 164
4.2.6 Recap of Proposition-Level Semantic Enhancements 165
4.3 Multicomponent Entities Recorded as Lexical Constructions 165
4.3.1 Semantically Null Components of Constructions 169
4.3.2 Typical Uses of Null-Semming 169
4.3.3 Modification of Null-Semmed Constituents 170
4.3.4 Utterance-Level Constructions 172
4.3.5 Additional Knowledge Representation Requirements 173
4.3.6 Recap of Constructions 175
4.4 Indirect Speech Acts, Lexicalized 175
4.5 Nominal Compounds, Lexicalized 176
4.6 Metaphors, Lexicalized 180
4.6.1 Past Work on Metaphor 180
4.6.2 Conventional Metaphors 182
4.6.3 Copular Metaphors 184
4.6.4 Recap of Metaphors 185
4.7 Metonymies, Lexicalized 185
4.8 Ellipsis 186
4.8.1 Verb Phrase Ellipsis 186
4.8.2 Verb Phrase Ellipsis Constructions 187
4.8.3 Event Ellipsis: Aspectual + NPoBJECT 189
4.8.4 Event Ellipsis: Lexically Idiosyncratic 189
4.8.5 Event Ellipsis: Conditions of Change 190
4.8.6 Gapping 192
4.8.7 Head Noun Ellipsis 192
4.8.8 Recap of Ellipsis 192
4.9 Fragmentary Utterances 193
4.10 Nonselection of Optional Direct Objects 193
4.11 Unknown Words 193
4.11.1 Completely Unknown Words 194
4.11.2 Known Words in a Different Part of Speech 196
4.12 Wrapping Up Basic Semantic Analysis 198
4.13 Further Exploration 199
5 Basic Coreference Resolution 201
5.1 A Nontechnical Introduction to Reference Resolution 201
5.1.1 Definitions 202
5.1.2 An Example-Based introduction 203
5.1.3 A Dozen Challenges 205
5.1.4 Special Considerations about Ellipsis 210
5.1.5 Wrapping Up the Introduction 211
5.2 Personal Pronouns 212
5.2.1 Resolving Personal Pronouns Using an Externally Developed Engine 213
5.2.2 Resolving Personal Pronouns Using Lexico-Syntactic Constructions 213
5.2.3 Semantically Vetting Hypothesized Pronominal Coreferences 215
5.2.4 Recap of Resolving Personal Pronouns during Basic Coreference Resolution 217
5.3 Pronominal Broad Referring Expressions 217
5.3.1 Resolving Pronominal Broad RefExes Using Constructions 218
5.3.2 Resolving Pronominal Broad RefExes in Syntactically Simple Contexts 219
5.3.3 Resolving Pronominal Broad RefExes Indicating Things That Must Stop 221
5.3.4 Resolving Pronominal Broad RefExes Using the Meaning of Predicate Nominals 223
5.3.5 Resolving Pronominal Broad RefExes Using Selectional Constraints 224
5.3.6 Recap of Resolving Pronominal Broad RefExes 226
5.4 Definite Descriptions 227
5.4.1 Definite Description Processing So Far: A Refresher 227
5.4.2 Definite Description Processing at This Stage 228
5.4.2.1 Rejecting coreference links with property value conflicts 228
5.4.2.2 Running reference-resolution meaning procedures listed in lexical senses 228
5.4.2.3 Establishing that a sponsor is not needed 229
5.4.2.4 Identifying bridging references 229
5.4.2.5 Creating sets as sponsors for plural definite descriptions 231
5.4.2.6 Identifying sponsors that are hypernyms or hyponyms of definite descriptions 232
5.4.3 Definite Description Processing Awaiting Situational Reasoning 233
5.4.4 Recap of Definite Description Processing at This Stage 233
5.5 Anaphoric Event Coreference 234
5.5.1 What is the Verbal/EVENT Head of the Sponsor? 235
5.5.2 Is There Instance or Type Coreference between the Events? 237
5.5.3 Is There Instance or Type Coreference between Objects in the VPs? 238
5.5.4 Should Adjuncts in the Sponsor Clause Be Included in, or Excluded from, the Resolution? 239
5.5.5 Should Modal and Other Scopers Be Included in, or Excluded from, the Resolution? 239
5.5.6 Recap of Anaphoric Event Coreference 240
5.6 Other Elided and Underspecified Events 241
5.7 Coreferential Events Expressed by Verbs 242
5.8 Further Exploration 244
6 Extended Semantic Analysis 247
6.1 Addressing Residual Ambiguities 248
6.1.1 The Objects Are Linked by a Primitive Property 248
6.1.2 The Objects Are Case Role Fillers of the Same Event 249
6.1.3 The Objects Are Linked by an Ontologically Decomposable Property 249
6.1.4 The Objects Are Clustered Using a Vague Property 251
6.1.5 The Objects Are Linked by a Short Ontological Path That Is Computed Dynamically 252
6.1.6 Reasoning by Analogy Using the TMR Repository 252
6.1.7 Recap of Methods to Address Residual Ambiguity 254
6.2 Addressing Incongruities 254
6.2.1 Metonymy 254
6.2.2 Preposition Swapping 256
6.2.3 Idiomatic Creativity 257
6.2.3.1 Detecting creative idiom use 258
6.2.3.2 Semantically analyzing creative idiom use 261
6.2.4 Indirect Modification Computed Dynamically 262
6.2.5 Recap of Treatable Types of Incongruities 264
6.3 Addressing Underspecification 264
6.3.1 Nominal Compounds Not Covered by Lexical Senses 264
6.3.2 Missing Values in Events of Change 270
6.3.3 Ungrounded and Underspecified Comparisons 270
6.3.4 Recap of Treatable Types of Underspecification 279
6.4 Incorporating Fragments into the Discourse Meaning 279
6.5 Further Exploration 284
7 Situational Reasoning 285
7.1 The OntoAgent Cognitive Architecture 286
7.2 Fractured Syntax 287
7.3 Residual Lexical Ambiguity: Domain-Based Preferences 290
7.4 Residual Speech Act Ambiguity 290
7.5 Underspecified Known Expressions 291
7.6 Underspecified Unknown Word Analysis 291
7.7 Situational Reference 292
7.7.1 Vetting Previously Identified Linguistic Sponsors for RefExes 293
7.7.2 Identifying Sponsors for Remaining RefExes 296
7.7.3 Anchoring the TMRs Associated with All RefExes in Memory 297
7.8 Residual Hidden Meanings 297
7.9 Learning by Reading 299
8 Agent Applications: The Rationale for Deep, Integrated NLU 301
8.1 The Maryland Virtual Patient System 301
8.1.1 Modeling Physiology 304
8.1.2 An Example: The Disease Model for GERD 306
8.13 Modeling Cognition 311
8.1.3.1 Learning new words and concepts through language interaction 312
8.1.3.2 Making decisions about action 313
8.1.4 An Example System Run 317
8.1.5 Visualizing Disease Models 320
8.1.5.1 Authoring instances of virtual patients 320
8.1.5.2 The knowledge about tests and interventions 321
8.1.5.3 Traces of system functioning 324
8.1.6 To What Extent Can MVP-Style Models Be Learned from Texts? 324
8.1.7 To What Extent Can Cognitive Models Be A Automatically Elicited from People? 328
8.2 A Clinician's Assistant for Flagging Cognitive Biases 331
8.2.1 Memory Support for Bias Avoidance 333
8.2.2 Detecting and Flagging Clinician Biases 335
8.2.3 Detecting and Flagging Patient Biases 339
8.3 LEIAs in Robotics 343
8.4 The Take-Home Message about Agent Applications 347
9 Measuring Progress 349
9.1 Evaluation Options-and Why the Standard Ones Don't Fit 350
9.2 Five Component-Level Evaluation Experiments 354
9.2.1 Nominal Compounding 355
9.2.2 Multiword Expressions 358
9.2.3 Lexical Disambiguation and the Establishment of the Semantic Dependency Structure 362
9.2.4 Difficult Referring Expressions 365
9.2.5 Verb Phrase Ellipsis 366
9.3 Holistic Evaluations 369
9.4 Final Thoughts 381
Epilogue 383
Notes 385
References 397
Index 415
