Text Genres and Registers: The Computation of Linguistic Features


Text Genres and Registers: The Computation of Linguistic Features (2015) .. by Chengyu Alex Fang, etc


Contents

1 Introduction … 1
1.1 The Corpus as a Model of Linguistic Use … 2
1.2 The Internal and External Dimensions in the Corpus … 3
1.3 The Predictive Power of the Corpus … 5
1.4 Genres and Registers … 6
1.5 Linguistic Variation across Genres and Registers … 9

2 Language Resources … 11
2.1 General Corpora … 11
2.1.1 The Brown Corpus and the Brown Family … 11
2.1.2 The International Corpus of English (ICE) Family … 14
2.1.3 BNC and ANC… 17
2.2 Specialised Collections … 22
2.2.1 Wall Street Journal … 22
2.2.2 PubMed … 23
2.3 Lexical Sources … 24
2.3.1 WordNet … 24
2.3.2 FrameNet … 24

3 Corpus Annotation and Usable Linguistic Features … 27
3.1 Textual Annotation … 28
3.2 Grammatical Annotation … 29
3.2.1 The LOB Tagset … 30
3.2.2 The ICE Tagset … 32
3.2.3 A Comparison of LOB and ICE … 35
3.3 Syntactic Annotation … 39
3.3.1 The Penn Treebank Scheme … 40
3.3.2 The ICE Parsing Scheme … 42
3.3.3 Summary … 44
3.4 Dialogue Act Annotation … 44
3.4.1 Notable DA Schemes … 47
3.4.2 ISO DA Scheme … 49
3.5 Machine Learning and Linguistic Features … 51
3.5.1 Machine Learning and Text Classification … 51
3.5.2 Weka … 53

4 Etymological Features across Genres and Registers … 55
4.1 Research Background … 55
4.2 Resources … 57
4.2.1 Corpus Resource … 57
4.2.2 Lexical Resource … 58
4.2.3 Reference Lists … 59
4.3 Investigation of Text Categories … 60
4.3.1 Descriptive Statistics … 60
4.3.2 Borrowed Words and Text Categories … 61
4.3.3 Summary … 64
4.4 Investigation of Subject Domains … 65
4.4.1 Creation of a Sub-corpus … 66
4.4.2 Descriptive Statistics … 66
4.4.3 Borrowed Words and Domains … 67
4.4.4 Summary … 69
4.5 Conclusion… 70

5 Part-of-Speech Tags and ICE Text Classification … 71
5.1 Research Background … 71
5.2 Methodology … 72
5.2.1 Experimental Setup … 72
5.2.2 Corpus Resources … 73
5.2.3 Machine-Learning Tools … 74
5.3 Feature Sets … 74
5.3.1 Fine-Grained POS Tags (F-POS) … 74
5.3.2 BOW … 75
5.3.3 Impoverished Tags (I-POS) … 75
5.4 Experimental Results … 76
5.4.1 Results Obtained From NB Classifier … 76
5.4.2 Results Obtained from NB-MN Classifier … 77
5.4.3 Discussion … 80
5.5 Conclusion… 82

6 Verbs and Text Classification … 83
6.1 Transitivity Type and Text Categories … 83
6.1.1 The Distribution of Lexical Verbs … 84
6.1.2 The Distribution of Verb Transitivity Types … 88
6.1.3 Conclusion… 95
6.2 Infinitive Verbs and Text Categories … 97
6.2.1 The Overall Distribution of Infinitives … 99
6.2.2 Aux Infinitives … 100
6.2.3 Bare Infinitives … 103
6.2.4 To-Infinitives … 105
6.2.5 For/to-Infinitives … 109
6.2.6 Summary and Conclusion … 115

7 Adjectives and Text Categories … 117
7.1 Adjective and Formality … 117
7.1.1 Research Background… 117
7.1.2 Methodology … 118
7.1.3 Adjective Use Across Text Categories … 119
7.1.4 Adjective Density and Automatic Text Classification … 124
7.1.5 Conclusion… 126
7.2 Adjective Phrase (AJP) and Subject Domains … 127
7.2.1 Corpus Resource … 127
7.2.2 Investigation of Adjective Use … 131
7.2.3 Conclusion… 133

8 Adverbial Clauses across Text Categories and Registers … 135
8.1 Adverbial Clauses Across Speech and Writing … 136
8.1.1 Adverbial Clauses Across Spontaneous and Prepared Speech … 137
8.1.2 Adverbial Clauses Across Timed and Untimed Essays … 138
8.2 Frequency Distribution of Adverbial Subordinators … 139
8.3 Discussions and Conclusion … 141
9 Coordination across Modes, Genres and Registers … 143
9.1 Methodology and Corpus Data … 150
9.2 The Distribution of Coordinators … 153
9.3 Syntactic Categories of Coordination Conjoins … 156
9.4 Syntactic Functions of Coordination … 160
9.5 Conclusion… 165

10 Semantic Features and Authorship Attribution … 167
10.1 Corpus Annotated with Ontological Knowledge … 171
10.2 Selection and Evaluation of Stylistic Features … 175
10.3 Discussions and Conclusion … 180

11 Pragmatics and Dialogue Acts … 183
11.1 Corpus Resource … 184
11.2 Related Research on the SWBD-DAMSL Scheme … 188
11.3 Methodology … 190
11.3.1 Machine Learning Techniques … 190
11.3.2 Data Preprocessing … 190
11.3.3 Research Questions … 191
11.4 Classification Results … 191
11.5 Qualitative Analysis … 194
11.5.1 Hedge … 194
11.5.2 Statement-non-Opinion Vs. Statement-Opinion … 203
11.5.3 Acknowledge (Backchannel) … 206
11.6 Conclusions … 214

12 The Future … 217

Appendix A: A List of ICE Part-of-Speech Tags … 221
Appendix B: A List of LOB Part-of-Speech Tags … 229
Appendix C: A List of Penn Treebank Part-of-Speech Tags … 233
Appendix D: A List of ICE Parsing Symbols… 235
Appendix E: A List of Penn Treebank Parsing Symbols … 237
Appendix F: A List of Adverbial Subordinators in Speech … 239
Appendix G: A List of Adverbial Subordinators in Writing … 243

Bibliography … 245
Index … 261

(Visited 25 times, 1 visits today)