The Discriminative Lexicon: Theory, Implementation in the Julia Package JudiLing, and Applications

The 'Discriminative Lexicon Model' is a new theory of how we process words, which moves radically away from most standard theories of morphology. This book introduces the Discriminative Lexicon from both a practical and a theoretical perspective. The first half explains the basic theory and the main parts of 'JudiLing', the Julia package implementing the theory. This is complimented by theory boxes introducing the core concepts underlying the model, such as Matrix Multiplication and the Rescorla-Wagner learning rule. The second half provides a series of case studies spanning languages as diverse as Maltese, Biblical Hebrew, Dutch, Navajo, Estonian and French, as well as multilingual settings. It also shows how behavioural data like lexical decision reaction times, acoustic durations or tongue movements can be modelled. These are accompanied by practice exercises. It is essential reading for researchers and students in a wide range of linguistic fields, including phonetics and computational linguistics.

Table of Contents:
1. Introduction; 2. Julia fundamentals; 3. The dataset; 4. Representing words' forms'; 5. Representing words' meanings'; 6. Computing comprehension and production mappings; 7. Evaluating mapping accuracy; 8. Producing forms; 9. Outputting modelling results; 10. Cross-validation; 11. Deep discriminative learning in JudiLing; 12. Worked examples – internal validation; 13. Predicting behavioural data with the DLM; 14. Worked examples – external validation; 15. No escape from morphemes?; 16. Further topics; 17. Concluding remarks.

About the Author :
Maria Heitmeier did her Ph.D. at the Eberhard Karls University, applying computational modeling to explore how humans process words. She currently works as a data scientist. Yu-Ying Chuang is Assistant Professor at National Taiwan Normal University. Her work explores phonetic variations in speech production and perception. Harald Baayen is Professor of quantitative linguistics at the University of Tuebingen. He is interested in how we learn to understand and produce the words of our language and uses statistical and error-driven computational modelling to understand how phonetic detail and lexical semantics interlink.