The Library
Functional data analysis in phonetics
Tools
Hadjipantelis, Pantelis-Zenon (2013) Functional data analysis in phonetics. PhD thesis, University of Warwick.
|
PDF
WRAP_THESIS_Hadjipantelis_2013.pdf - Submitted Version - Requires a PDF viewer. Download (4Mb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b2732211~S1
Abstract
The study of speech sounds has established itself as a distinct area of research, namely Phonetics. This is because speech production is a complex phenomenon mediated by the interaction of multiple components of a linguistic and non-linguistic nature. To investigate such phenomena, this thesis employs a Functional Data Analysis framework where speech segments are viewed as functions. FDA treats functions as its fundamental unit of analysis; the thesis takes advantage of this, both in conceptual as well as practical terms, achieving theoretical coherence as well as statistical robustness in its insights. The main techniques employed in this work are: Functional principal components analysis, Functional mixed-effects regression models and phylogenetic Gaussian process regression for functional data. As it will be shown, these techniques allow for complementary analyses of linguistic data. The thesis presents a series of novel applications of functional data analysis in Phonetics. Firstly, it investigates the influence linguistic information carries on the speech intonation patterns. It provides these insights through an analysis combining FPCA with a series of mixed effect models, through which meaningful categorical prototypes are built. Secondly, the interplay of phase and amplitude variation in functional phonetic data is investigated. A multivariate mixed effects framework is developed for jointly analysing phase and amplitude information contained in phonetic data. Lastly, the phylogenetic associations between languages within a multi-language phonetic corpus are analysed. Utilizing a small subset of related Romance languages, a phylogenetic investigation of the words' spectrograms (functional objects defined over two continua simultaneously) is conducted to showcase a proof-of-concept experiment allowing the interconnection between FDA and Evolutionary Linguistics.
Item Type: | Thesis (PhD) | ||||
---|---|---|---|---|---|
Subjects: | P Language and Literature > P Philology. Linguistics Q Science > QA Mathematics |
||||
Library of Congress Subject Headings (LCSH): | Phonetics -- Mathematical models, Mathematical statistics | ||||
Official Date: | December 2013 | ||||
Dates: |
|
||||
Institution: | University of Warwick | ||||
Theses Department: | Department of Statistics | ||||
Thesis Type: | PhD | ||||
Publication Status: | Unpublished | ||||
Supervisor(s)/Advisor: | Aston, John A. D. ; Evans, Jonathan P. | ||||
Extent: | x, 198 leaves : charts | ||||
Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
View Item |
Downloads
Downloads per month over past year