Syntactic wordclass tagging / edited by Hans van Halteren.

Format:

Book

Contributor:

Halteren, Hans van.

Alumni and Friends Memorial Book Fund.

Series:

Text, speech, and language technology ; v. 9.

Text, speech, and language technology ; v. 9

Language:

English

Subjects (All):

Grammar, Comparative and general--Morphosyntax.

Grammar, Comparative and general.

Parts of speech.

Computational linguistics.

Physical Description:

xvii, 334 pages : illustrations ; 25 cm.

Place of Publication:

Dordrecht ; Boston : Kluwer Academic Publishers, [1999]

Contents:

Part I The User's View

1 Orientation / Atro Voutilainen 3

1.1 Morphosyntactic tags 3

1.2 Automatic tagging 6

2 A Short History of Tagging / Atro Voutilainen 9

2.1 Approaches to wordclass tagging 9

2.2 Pioneering work 10

2.3 The breakthrough of data-driven methods 11

2.3.1 N-gram taggers 12

2.3.2 Data-driven local rules 13

2.4 Recent work in the data-driven approach 14

2.4.1 Hidden Markov Models 14

2.4.2 Recent work on data-driven local rules 16

2.4.3 Neural taggers 16

2.4.4 Case-based taggers 17

2.4.5 Combined data-driven taggers 17

2.5 Recent work in the linguistic approach 17

2.5.1 English Constraint Grammar 18

2.5.2 A rule-based tagger of Turkish 18

2.5.3 A finite-state tagger of French 19

2.5.4 A syntax-based tagger of English 19

2.6 The current situation 19

3 The Use of Tagging / Geoffrey Leech, Nicholas Smith 23

3.2 Tagging in corpus linguistics 24

3.2.1 Adding further annotations 26

3.2.2 Information extraction 28

3.3 Practical applications 31

3.3.1 Uses of tagging software 31

3.3.2 Uses of tagged text 33

4 Tagsets / Jan Cloeren 37

4.2 Information contents of the tags in the tagset 37

4.2.1 Morphosyntactic tags 38

4.2.2 Syntactic tags 40

4.2.3 Semantic and discourse tags 41

4.2.4 Distributional similarity tags 42

4.3 Special problems in the application of tagsets 44

4.3.1 Multi-unit tokens and multi-token units 44

4.3.2 Underspecification and ambiguity 46

4.4.1 Class and feature value names 49

4.4.2 Structure of tags 50

4.4.3 Positioning of tags 51

4.4.4 SGML/TEI guidelines for tags 51

5 Standards for Tagsets / Geoffrey Leech, Andrew Wilson 55

5.2 Recommendations for morphosyntactic (wordclass) categories 58

5.2.1 Reasonable goals for standardization 58

5.2.2 Word categories: tagset guidelines 61

5.3 Intermediate Tagset 70

5.3.1 Basic Structure 70

5.3.2 Underspecification 71

5.3.3 Example tagsets 72

6 Performance of Taggers / Hans van Halteren 81

6.2 Performance measures 81

6.2.1 Definitions of measures 82

6.2.2 Usefulness of measures 83

6.3 Performance measurements 86

6.3.1 Experimental setup 86

6.3.2 Effects of the tagset 87

6.3.3 Effects of the method of comparison 89

6.3.4 Effects of choice of tokens measured 90

6.3.5 Effects of separation of test and training material 91

6.3.6 Effects of representativity of test material 94

7 Selection and Operation of Taggers / Hans van Halteren 95

7.2 Selection of a tagger 95

7.2.1 Tagset 96

7.2.2 Documentation 96

7.2.3 The tagging process 97

7.2.4 Performance 98

7.2.5 Combining the factors 98

7.3 User interaction 99

7.3.1 Tokenization 100

7.3.2 Classification of unknown tokens 101

7.3.3 Selection of the contextually appropriate tag 101

7.3.4 Post-processing of tagged text 102

Appendix NOT an inventory of taggers 103

Part II The Implementer's View

8 Automatic Taggers: An Introduction / Hans van Halteren, Atro Voutilainen 109

8.1 General architecture 109

8.1.1 Tokenization 110

8.1.2 Assignment of potential tags 110

8.1.3 Determination of the most likely tag 110

8.2 Corpus resources 110

8.2.1 Form of corpus resources 111

8.2.2 Size of corpus resources 113

8.2.3 Creation of corpus resources 114

9 Tokenization / Gregory Grefenstette 117

9.2 Regular expressions 119

9.2.2 Regular expression tools LEX and AWK 121

9.2.3 An example of a tokenizer 121

9.3 Ambiguity in tokenization 125

9.3.1 Splitting graphic tokens 125

9.3.2 Combining graphic tokens 132

10 Lexicons for Tagging / Anne Schiller, Lauri Karttunen 135

10.2 Morphology-based lexicons 137

10.2.1 Direct mapping 140

10.2.2 Merging morphological classes 141

10.2.3 Refining morphological classes 141

10.2.4 Adding residual wordclasses 143

10.3 Corpus-based lexicons 144

10.3.1 Enlarged Training Corpus 146

10.3.2 External Lexical Resources 146

11 Standardization in the Lexicon / Monica Monachini, Nicoletta Calzolari 149

11.1 The initiative for standardization 149

11.2 Interdependence between lexicon and corpus 151

11.2.1 Lexical encoding vs. corpus tagsets 151

11.2.2 Tagsets as collapsed feature specifications 152

11.2.3 Multi-linguality 152

11.2.4 Lexicon specifications as an interface between tagsets 153

11.3 The EAGLES proposal for morphosyntactic encoding 156

11.3.1 Methodology of standardization 157

11.3.2 The proposal 159

11.4 Instantiation in different languages 161

11.5 Guidelines for the validation phase 164

11.5.1 Values pertinent to a given language 165

11.5.2 Logic relationships between values 165

11.5.3 Constraints in the application of attributes and values 167

11.5.4 Semantics of the PoS 168

11.5.5 Semantics of the features 170

11.6 Application in EU projects 171

11.6.1 Multext 171

11.6.2 Parole 172

11.6.3 Coverage with respect to languages, users and applications 173

12 Morphological Analysis / Kemal Oflazer 175

12.2 Morphology 177

12.2.1 Types of morphology 177

12.2.2 Types of morphological combination 178

12.2.3 Computational morphology 179

12.3 Two-level morphology 180

12.3.1 The morphographemic component 181

12.3.2 The morphotactics component 186

12.3.3 Development tools 190

12.3.4 Developing a Morphological Analyser 193

12.4 A morphological analyser for Turkish 194

12.4.1 Requirements 195

12.4.2 System architecture 199

12.4.3 The morphographemic transducer: T[subscript is-lx] 202

12.4.4 The morphotactics transducer: T[subscript lx-if] 203

13 Tagging Unknown Words / Eric Brill 207

13.2 Behaviour of unknown words 207

13.3 Dealing with unknown words 209

13.4 Unknown words in case-based tagging 211

13.5 Unknown words in transformation-based tagging 212

13.6 Lexicon extrapolation 215

14 Hand-Crafted Rules / Atro Voutilainen 217

14.2 Comparison of paradigms 218

14.3 Rule formalism 219

14.3.2 Operations 220

14.3.3 Targets 221

14.3.4 Context conditions 221

14.3.5 Sample rules 222

14.3.6 Some facts about a large grammar 224

14.4 Writing a disambiguation grammar 226

14.4.1 A sample session 227

14.4.2 Experiences with novices: NorFa'95 CG 'competition' 240

14.5 General observations 242

14.6 Remaining ambiguity 243

14.6.1 Using statistical models 244

14.6.2 Using collocational information 244

14.6.3 Using a syntactic parser 245

14.6.4 Using observed local regularities 245

15 Corpus-Based Rules / Eric Brill 247

15.2 Learning rules 248

15.3 Parser-based wordclass disambiguation 249

15.4 Transformation-based learning 251

15.5 N-best wordclass tagging 256

15.6 Unsupervised learning 258

15.7 Issues of portability 261

16 Hidden Markov Models / Marc El-Beze, Bernard Merialdo 263

16.2 HMMs in general 264

16.2.2 An example 265

16.2.3 Choosing the underlying topology 267

16.2.4 Training 268

16.2.5 Decoding 269

16.3 HMMs for wordclass tagging 271

16.3.1 The structure of the model 271

16.3.2 Choice of tagset when using HMMs 273

16.4 Training HMMs for tagging 274

16.4.1 Training on tagged text 274

16.4.2 Smoothing the triclass model 276

16.4.3 Training HMM taggers with Baum-Welch 278

16.5 Tagging with HMMs 280

16.5.1 Using the Viterbi algorithm 280

16.5.2 Other forms of decoding 282

16.6 Combining different linguistic levels 282

16.6.1 Using wordclasses in word models 282

16.6.2 Using lemma models in wordclass tagging 284

17 Machine Learning Approaches / Walter Daelemans 285

17.2 Inductive learning from examples 287

17.2.1 Concepts 287

17.2.2 Classification of learning methods 288

17.2.3 Performance evaluations 290

17.2.4 Overview of methods 290

17.3 Case-based learning 291

17.3.1 Algorithm 292

17.3.2 Case-based tagging 293

17.3.3 Evaluation 296

17.4 Decision tree induction 297

17.4.1 Algorithm 297

17.4.2 Decision tree tagging 298

17.4.3 Evaluation 300

17.5 Neural network methods 300

17.5.1 Algorithm 301

17.5.2 Neural network tagging 302

17.5.3 Evaluation 303

Appendix A Example tagsets 305

A.1 The Brown Corpus tagset 305

A.2 The Penn Treebank tagset 307

A.3 The EngCG tagset 309.

Notes:

Includes bibliographical references (pages 311-326) and index.

Local Notes:

Acquired for the Penn Libraries with assistance from the Alumni and Friends Memorial Book Fund.

ISBN:

0792358961

OCLC:

41674228