Eden, S. Elizabeth; (2018) Measuring phonological distance between languages. Doctoral thesis (Ph.D), UCL (University College London).

Preview	Text (Thesis) Eden_10058348_thesis.pdf Download (1MB) | Preview
	Archive (Supplementary files 1) Eden_10058348_thesis_supplementary_1.zip Download (4kB)
	Archive (Supplementary files 2) Eden_10058348_thesis_supplementary_2.zip Download (21MB)
	Archive (Supplementary files 3) Eden_10058348_thesis_supplementary_3.zip Download (25MB)

Abstract

Three independent approaches to measuring cross-language phonological distance are pursued in this thesis: exploiting phonological typological parameters; measuring the cross-entropy of phonologically transcribed texts; and measuring the phonetic similarity of non-word nativisations by speakers from different language backgrounds. Firstly, a set of freely accessible online tools are presented to aid in establishing parametric values for syllable structure and phoneme inventory in different languages. The tools allow researchers to make differing analytical and observational choices and compare the results. These tools are applied to 16 languages, and correspondence between the resulting parameter values is used as a measure of phonological distance. Secondly, the computational technique of cross-entropy measurement is applied to texts from seven languages, transcribed in four different ways: a phonemic IPA transcription; with Elements; and with two sets of binary distinctive features in the SPE tradition. This technique results in consistently replicable rankings of phonological similarity for each transcription system. It is sensitive to differences in transcription systems. It can be used to probe the consequences for information transfer of the choices made in devising a representational system. Thirdly, participants from different language backgrounds are presented with non-words covering the vowel space, and asked to nativise them. The accent distance metric ACCDIST is applied to the resulting words. A profile of how each speaker’s productions cluster in the vowel space is produced, and ACCDIST measures the similarity of these profiles. Averaging across speakers with a shared native language produces a measure of similarity between language profiles. Each of these three approaches delivers a quantitative measure of phonological similarity between individual languages. They are each sensitive to different analytical choices, and require different types and quantities of input data, and so can complement each other. This thesis provides a proof-of-concept for methods which are both internally consistent and falsifiable.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item