基于语法自举原理的儿童动词学习模型研究

Chapter 1 Introduction1

Chapter 2 Language Acquisition3

2.1 What is Language Acquisition3

2.2 The Course of Language Acquisition4

2.3 Language Acquisition Theories5

2.3.1 Universal Grammar5

2.3.2 Learnability Theory6

2.4 Acquisition Algorithm7

2.4.1 Using Correlation Matrix7

2.4.2 Identifying Prosody7

2.4.3 Using Context Information8

Chapter 3 Verb Acquisition Theories10

3.1 What Kind of Information Children Learn10

3.1.1 Subcategorization Frame10

3.1.2 Thematic Role11

3.2 Semantic Bootstrapping VS. Syntactic Bootstrapping13

3.2.1 Syntactic/Semantic Linking Rules13

3.2.2 Semantic Bootstrapping14

3.2.3 Syntactic Bootstrapping14

Chapter 4 Algorithm of Children's Verb Acquisition17

4.1 Using Verb Prototypes to Guide Children's Verb SF Acquisition17

4.1.1 Constructing Verb Prototypes17

4.1.2 Argument Structures of Verb Prototypes19

4.1.3 How to Acquire SF for a Novel Verb20

4.2 Thematic Role Acquisition20

4.2.1 Thematic Role Identification Based on Syntactic Bootstrapping21

4.2.2 Heuristics for Determining Thematic Roles21

4.2.3 Using Multiple Subcategorization Frames to Help Assign Thematic Roles23

4.2.4 How to Acquire Thematic Roles24

Chapter 5 Verb Learning Model26

5.1 Introduction27

5.1.1 Child Parser27

5.1.2 KDB and IDB30

5.1.3 CM and AM30

5.2 Abstract Verb Lexicon31

5.3 How Does the Model Work32

5.3.1 Interface of the Model32

5.3.2 An Example for Processing a Sentence Input33

Chapter 6 Experiments and Results35

6.1 Experimental Data35

6.1.1 Introduction to CHILDES Database36

6.1.2 Maternal Corpus37

6.1.3 Child Corpus38

6.2 Experimental Results39

6.2.1 Results of Maternal Corpus39

6.2.2 Results of Child Corpus40

6.3 Discussion40

6.3.1 Problems in SF Acquisition41

6.3.2 Problems in Thematic Role Acquisition41

6.3.3 Hypothesis of Verb Prototype42

Chapter 7 Conclusion of the Child Verb Learning Model43

Chapter 8 An Application System of the Child Verb Learning Model45

8.1 Introduction45

8.2 How to Acquire Thematic Roles46

8.2.1 Cues Based on the Child Verb Learning Model46

8.2.2 Using DYN and CON to Identify the Thematic Roles for Verb Arguments47

8.2.3 Decision Criteria of Thematic Role48

8.3 Experiments and Results49

8.3.1 Optimum Examples49

8.3.2 Experiments50

8.3.3 Results51

8.4 Discussion51

8.4.1 The Use of Optimum Examples52

8.4.2 Improving the Algorithm52

8.5 Conclusion53

Chapter 9 Conclusions 54

9.1 Contributions54

9.2 Future Work56

References 57

Appendix A Data used in the experiments63

A.1 Example Sentences in Maternal Corpus63

A.2 Example Sentences in Child Corpus67

Appendix B Experimental Results71

B.1 Example Results from Experiment Based on Maternal Corpus71

B.2 Example Results from Experiment Based on Child Corpus78

本书既是依据乔姆斯基和平克等学者的认知学及语言学的研究，对建立4-5岁儿童学习英语动词的计算机模型的研究。书中，首先对语言心理学(Psycholinguistics)及认知学（Cognitive science）中关于幼儿动词学习的各种理论进行归纳总结，并对母子对话对语义语法学习的影响进行研究。然后，利用Syntactic Bootstrapping、Agent Bias原理等认知心理学及语言心理学的研究成果，来对自然语言处理研究中的动词语义语法知识的自动学习进行研究应用，建立了一个能够自动学习语法语义信息的英语动词模型。