Turkish Natural Language Processing

ISBN-13: 9783030079499 / Angielski / Miękka / 2019 / 355 str.

Kemal Oflazer; Murat Saraclar

Spis treści

1 Turkish and its Challenges for Language and Speech Processing . . . . 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Turkish Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Constituent Order and Morphology-Syntax Interface . . . . . . . . . . . . 7

1.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5 State-of-the-art Tools and Resources for Turkish . . . . . . . . . . . . . . . 15

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2 Morphological Processing for Turkish . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.2 Overview of Turkish Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.3 Morphophonology and Morphographemics . . . . . . . . . . . . . . . . . . . . 23

2.4 Root Lexicons and Morphotactics . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.4.1 Representational Convention . . . . . . . . . . . . . . . . . . . . . . . . 28

2.4.2 Nominal Morphotactics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.4.3 Verbal Morphotactics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.4.4 Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.4.5 Examples of Morphological Analyses . . . . . . . . . . . . . . . . 32

2.5 The Architecture of the Turkish Morphological Processor . . . . . . . . 34

2.6 Processing Real Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.6.1 Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.6.2 Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.6.3 Foreign Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.6.4 Unknown Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.7 Multiword Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.7.1 Lexicalized Collocations . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.7.2 Semi-lexicalized Collocations . . . . . . . . . . . . . . . . . . . . . . . 38

2.7.3 Non-lexicalized Collocations . . . . . . . . . . . . . . . . . . . . . . . . 40

2.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3 Morphological Disambiguation for Turkish . . . . . . . . . . . . . . . . . . . . . . 53

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.2 Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.3 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.3.1 Rule-based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.3.2 Learning the Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.3.3 Models Based on Inflectional Group n-grams . . . . . . . . . . 59

3.3.4 Discriminative Methods for Disambiguation . . . . . . . . . . . 60

3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.4.1 Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.4.2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4 Language Modeling for Turkish Text and Speech Processing . . . . . . . 69

Ebru Arısoy and Murat Saraçlar

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.2 Language Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.3 Challenges in Statistical Language Modeling for Turkish . . . . . . . . 73

4.4 Sub-lexical Units for Statistical Language Modeling . . . . . . . . . . . . 75

4.4.1 Linguistic Sub-lexical Units . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.4.2 Statistical Sub-lexical Units . . . . . . . . . . . . . . . . . . . . . . . . . 77

4.5 Statistical Language Modeling for Turkish . . . . . . . . . . . . . . . . . . . . 78

4.5.1 Language Modeling with Linguistic Sub-lexical Units . . . 78

4.5.2 Statistical Sub-lexical Units – Morphs . . . . . . . . . . . . . . . . 81

4.6 Discriminative Language Modeling for Turkish . . . . . . . . . . . . . . . . 81

4.6.1 Discriminative Language Model . . . . . . . . . . . . . . . . . . . . . 82

4.6.2 Feature Sets for Turkish DLM . . . . . . . . . . . . . . . . . . . . . . . 83

4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5 Turkish Speech Recognition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

Ebru Arısoy and Murat Saraçlar

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.2 Foundations of Automatic Speech Recognition . . . . . . . . . . . . . . . . 96

5.3 Turkish Language Resources for ASR . . . . . . . . . . . . . . . . . . . . . . . . 100

5.3.1 Turkish Acoustic and Text Data . . . . . . . . . . . . . . . . . . . . . . 100

5.3.2 Linguistic Tools Used in Turkish ASR . . . . . . . . . . . . . . . . 105

5.4 Turkish ASR Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5.4.1 Newspaper Content Transcription System . . . . . . . . . . . . . 106

5.4.2 Turkish Broadcast News Transcription System . . . . . . . . . 109

5.4.3 LVCSR System for Call Center Conversations . . . . . . . . . 112

5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

6 Turkish Named Entity Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

Reyyan Yeniterzi, Gökhan Tür and Kemal Oflazer

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

6.2 NER on Turkish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

6.3 Task Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

6.3.1 Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

6.3.2 Evaluating NER Performance . . . . . . . . . . . . . . . . . . . . . . . 122

6.4 Domain and Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.4.1 Formal Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.4.2 Informal Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6.4.3 Challenges of Informal Texts for NER . . . . . . . . . . . . . . . . 126

6.5 Preprocessing for NER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

6.5.1 Tokenization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.5.2 Morphological Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.5.3 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

6.6 Approaches used in Turkish NER . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

6.6.1 Rule-based Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

6.6.2 Hybrid Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

6.6.3 Machine Learning Approaches . . . . . . . . . . . . . . . . . . . . . . 131

6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

7 Dependency Parsing of Turkish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

7.2 Dependency Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

7.3 Morphology and Dependency Relations in Turkish . . . . . . . . . . . . . 140

7.3.1 Dependency Relations in Turkish . . . . . . . . . . . . . . . . . . . . 143

7.4 An Incremental Data-driven Statistical Dependency Parsing

System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

7.4.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

7.4.2 Modeling Turkish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

7.4.3 Evaluation Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

7.5 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

7.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

8 Wide-coverage parsing, semantics and morphology . . . . . . . . . . . . . . . 157

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

8.2 Morphology and Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

8.3 Radical Lexicalization and Predicate-Argument Structure of

sub-lexical Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

8.4 Combinatory Categorial Grammar: CCG. . . . . . . . . . . . . . . . . . . . . . 162

8.5 The Turkish Categorial Lexicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

8.5.1 The Lexemic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

8.5.2 The Morphemic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

8.6 Parsing with Automatically Induced CCG Lexicons . . . . . . . . . . . . 172

8.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

9 Deep Parsing of Turkish with Lexical-Functional Grammar . . . . . . . . 179

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

9.2 Lexical-Functional Grammar and Xerox Linguistic Environment . 180

9.3 Inflectional Groups as First-class Syntactic Citizens . . . . . . . . . . . . 181

9.4 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

9.5 LFG Analyses of Various Linguistic Phenomena . . . . . . . . . . . . . . . 185

9.5.1 Noun Phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

9.5.2 Adjective Phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

9.5.3 Adverbial Phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

9.5.4 Postpositional Phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

9.5.5 Temporal Phrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

9.6 Sentential Derivations, Sentences and Free Constituent Order . . . . 189

9.6.1 Sentential Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

9.6.2 Sentences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

9.6.3 Handling Constituent Order Variations . . . . . . . . . . . . . . . . 195

9.7 Coordination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

9.8 Valency Alternations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

9.8.1 Causatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

9.8.2 Passives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

9.9 Non-canonical Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

9.10 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

9.10.1 Manual Test Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

9.10.2 Sentence Test Suite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

9.10.3 Noun Phrase Test Suite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

9.11 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

10 Statistical Machine Translation and Turkish . . . . . . . . . . . . . . . . . . . . . 213

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10.2 Handling Morphology in Statistical Machine Translation . . . . . . . . 215

10.3 The Morpheme Segmentation Approach . . . . . . . . . . . . . . . . . . . . . . 216

10.3.1 Experiments and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

10.3.2 Word Repair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

10.3.3 Sample Translations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

10.3.4 Observations on the Morpheme Segmentation Approach . 224

10.4 The Syntax-to-Morphology Mapping Approach . . . . . . . . . . . . . . . . 225

10.4.1 Mapping Source-side Syntax to Target-side Morphology . 226

10.4.2 Experimental Setup and Results . . . . . . . . . . . . . . . . . . . . . 230

10.4.3 Experiments with Constituent Reordering . . . . . . . . . . . . . 237

10.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241

11 Machine Translation Between Turkic Languages . . . . . . . . . . . . . . . . . 245

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

11.2 Turkic Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

11.2.1 Similarities and Differences of Turkic Languages . . . . . . . 246

11.3 Machine Translation between Turkic Languages . . . . . . . . . . . . . . . 250

11.3.1 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

11.3.2 Morphological Disambiguation . . . . . . . . . . . . . . . . . . . . . . 253

11.3.3 Morphological Feature Transfer . . . . . . . . . . . . . . . . . . . . . 254

11.3.4 Lexical Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254

11.3.5 Statistical Disambiguation Module . . . . . . . . . . . . . . . . . . . 256

11.3.6 Sentence Level Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

11.3.7 Morphological Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 258

11.4 Machine Translation Evaluation on Turkic Languages . . . . . . . . . . . 258

11.4.1 Root Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

11.4.2 Feasible Suffix Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260

11.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

12 Sentiment Analysis in Turkish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

12.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

12.3 Main Difficulties for Turkish Sentiment Analysis . . . . . . . . . . . . . . . 270

12.4 Practical Sentiment Analysis for Turkish . . . . . . . . . . . . . . . . . . . . . . 271

12.4.1 Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

12.4.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

12.5 Experimental Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

12.5.1 Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

12.5.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

12.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

13 The Turkish Treebank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

13.2 What information needs to be represented? . . . . . . . . . . . . . . . . . . . . 284

13.2.1 Representing Morphological Information . . . . . . . . . . . . . . 284

13.2.2 Representing Syntactic Relations . . . . . . . . . . . . . . . . . . . . 286

13.2.3 Example of a Treebank Sentence . . . . . . . . . . . . . . . . . . . . . 288

13.3 Evolution of the Turkish Treebank . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

13.3.1 The CoNLL Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

13.3.2 Branches of the Turkish Treebank . . . . . . . . . . . . . . . . . . . . 292

13.4 The ITU Web Treebank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293

13.5 The Annotation Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

13.6 The Turkish Universal Dependencies Treebank . . . . . . . . . . . . . . . . 296

13.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298

14 Linguistic corpora: A view from Turkish . . . . . . . . . . . . . . . . . . . . . . . . 301

14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

14.2 Brief History of Corpus Linguistics . . . . . . . . . . . . . . . . . . . . . . . . . . 302

14.3 Linguistic Corpora and Corpus Linguistics . . . . . . . . . . . . . . . . . . . . 304

14.4 Use of Corpora in Linguistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308

14.5 Turkish Linguistic Corpora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309

14.5.1 METU-Turkish Corpus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310

14.5.2 Turkish National Corpus (TNC) . . . . . . . . . . . . . . . . . . . . . 313

14.5.3 Spoken Turkish Corpus (STC) . . . . . . . . . . . . . . . . . . . . . . . 315

14.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321

15 Turkish Wordnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327

15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327

15.2 Basic Structure of Turkish Wordnet . . . . . . . . . . . . . . . . . . . . . . . . . . 328

15.2.1 Semantic Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328

15.2.2 Linking Wordnets to Each Other . . . . . . . . . . . . . . . . . . . . . 329

15.3 Design Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330

15.3.1 Merge vs. Expand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331

15.3.2 Parts-of-Speech, Definitions and Sense Numbers . . . . . . . 331

15.3.3 Lexical Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332

15.3.4 No Dangling Nodes or Relations . . . . . . . . . . . . . . . . . . . . . 332

15.3.5 Validating Semantic Relations . . . . . . . . . . . . . . . . . . . . . . . 333

15.4 The Development Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

15.4.1 First Set of Concepts (Subset I) . . . . . . . . . . . . . . . . . . . . . . 333

15.4.2 Extracting Semantic Relations from Monolingual

Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334

15.4.3 Second Set of Concepts (Subset II) . . . . . . . . . . . . . . . . . . . 336

15.4.4 Shifting to Princeton Wordnet 1.7.1 . . . . . . . . . . . . . . . . . . 337

15.4.5 Third Set of Concepts (Subset III) . . . . . . . . . . . . . . . . . . . . 338

15.4.6 Shifting to Princeton Wordnet 2.0 . . . . . . . . . . . . . . . . . . . . 338

15.4.7 Adding Balkanet-specific Concepts. . . . . . . . . . . . . . . . . . . 338

15.4.8 Final Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339

15.5 Current Status of Turkish Wordnet . . . . . . . . . . . . . . . . . . . . . . . . . . . 339

15.6 Quality Validation and Coverage Tests . . . . . . . . . . . . . . . . . . . . . . . . 340

15.7 Applications of Turkish Wordnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342

15.7.1 Capturing Semantic Relations through Morphology . . . . . 342

15.7.2 Turkish Wordnet in Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344

15.8 Conclusion and Directions for Future Work . . . . . . . . . . . . . . . . . . . 345

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345

16 Turkish Discourse Bank: Connectives and Their Configurations . . . . 349

16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349

16.2 The TDB Annotation Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351

16.2.1 Major Sources of Disagreements among Annotators . . . . 353

16.2.2 The Discourse Annotation Tool for Turkish . . . . . . . . . . . . 355

16.3 Connectives and Discourse Structure . . . . . . . . . . . . . . . . . . . . . . . . . 355

16.4 Discourse relation configurations in the TDB . . . . . . . . . . . . . . . . . . 356

16.4.1 Independent Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357

16.4.2 Full Embedding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358

16.4.3 Nested Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358

16.4.4 Shared Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360

16.4.5 Properly Contained Argument . . . . . . . . . . . . . . . . . . . . . . . 360

16.4.6 Properly Contained Relation . . . . . . . . . . . . . . . . . . . . . . . . 362

16.4.7 Partially Overlapping Arguments . . . . . . . . . . . . . . . . . . . . 362

16.4.8 Pure Crossing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364

16.5 Results and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366

O autorze

Kemal Oflazer received his Ph.D. in computer science from Carnegie Mellon University in Pittsburgh, USA, and his M.Sc. in computer science and B.Sc. in electrical and electronics engineering from Middle East Technical University, Ankara, Turkey. He is currently a faculty member at Carnegie Mellon University in Doha, Qatar, where he is also the Associate Dean for Research. He has held visiting positions at the Computing Research Laboratory at New Mexico State University, Las Cruces, USA and at the Language Technologies Institute, Carnegie Mellon University. Prior to joining CMU-Qatar, he worked at Sabancı University in Istanbul, Turkey (2000-2008) and Bilkent University in Ankara, Turkey (1989-2000). He has worked extensively on developing natural language processing techniques and resources for Turkish. Oflazer’s current research interests include statistical machine translation into morphologically complex languages, the use of NLP for language learning and machine learning for computational morphology. In addition, he was a member of the editorial boards of Computational Linguistics, the Journal of Artificial Intelligence Research, Machine Translation, and Research on Language and Computation and was a book review editor for Natural Language Engineering. He was a member of the nomination and advisory boards for EACL, and served as the program co-chair for ACL 2005, an area chair for COLING 2000, EACL 2003, ACL 2004, ACL 2012, and EMNLP 2013 and the organization committee co-chair for EMNLP 2014. Currently, he is an editorial board member of both Language Resources and Evaluation and Natural Language Engineering journals and is a member of the advisory board for “SpringerBriefs in Natural Language Processing”.

Murat Saraçlar received his B.Sc. degree in 1994 from the Electrical and Electronics Engineering Department at Bilkent University, Ankara, Turkey, his M.S.E. degree in 1997 and Ph.D. degree in 2001 from the Electrical and Computer Engineering Department at the Johns Hopkins University, Baltimore, USA. From 2000 to 2005, he was with the multimedia services department at AT&T Labs Research, and in 2005 joined the Electrical and Electronic Engineering Department of Boğaziçi University, Istanbul, Turkey, where he is currently a full professor. He was a visiting research scientist at Google Inc., New York, USA (2011-2012) and an academic visitor at IBM T.J. Watson Research Center (2012-2013). Saraçlar was awarded the AT&T Labs Research Excellence Award in 2002, the Turkish Academy of Sciences Young Scientist (TUBA-GEBIP) Award in 2009, and the IBM Faculty Award in 2010. He has published more than 100 articles in journals and conference proceedings. Furthermore, he served as an associate editor for IEEE Signal Processing Letters (2009-2012) and IEEE Transactions on Audio, Speech, and Language Processing (2012-2016). He was an editorial board member of Language Resources and Evaluation from 2012 to 2016, and is currently an editorial board member of Computer Speech and Language as well as a member of the IEEE Signal Processing Society Speech and Language Technical Committee (2007-2009, 2015-2018).

Blurb

This book brings together work on Turkish natural language and speech processing over the last 25 years, covering numerous fundamental tasks ranging from morphological processing and language modeling, to full-fledged deep parsing and machine translation, as well as computational resources developed along the way to enable most of this work. Owing to its complex morphology and free constituent order, Turkish has proved to be a fascinating language for natural language and speech processing research and applications.

After an overview of the aspects of Turkish that make it challenging for natural language and speech processing tasks, this book discusses in detail the main tasks and applications of Turkish natural language and speech processing. A compendium of the work on Turkish natural language and speech processing, it is a valuable reference for new researchers considering computational work on Turkish, as well as a one-stop resource for commercial and research institutions planning to develop applications for Turkish. It also serves as a blueprint for similar work on other Turkic languages such as Azeri, Turkmen and Uzbek.