Tools and Techniques for Corpus-based Analysis of Non-native Speech Patterns

Abstract

This study examines the problem of describing non-native speech from the perspective of corpus linguistics. Recent data storage and manipulation techniques have allowed for the development of rich phonological corpora, but few corpora have been developed or explored specifically to support accented speech research. This dissertation evaluates different theories used to explain the existence and structure of non-native speech. It then details the requirements for using phonological corpora to support accented speech research. In particular, it argues that narrow phonetic transcriptions of elicited speech can be used to identify and analyze non-native speech patterns. To easily evaluate a corpus of comparable phonetic transcriptions this dissertation discusses the use of a new set of tools for the rapid analysis of phonological corpora. It then details how these tools can be applied to identify a list of non-native speech patterns. Finally, it details how these new corpus-based studies of non-native speech can be used with newer approaches to phonological grammars that incorporate weighted constraints. This dissertation demonstrates that phonological corpora containing phonetic transcriptions can indeed be an incredible resource for research in the understanding of non-native speech patterns.