Studying dialects to understand human language

Abstract

This thesis investigates the study of dialect variations as a way to understand how humans might process speech. It evaluates some of the important research in dialect identification and draws conclusions about how their results can give insights into human speech processing. A study clustering dialects using k-means clustering is done. Self-organizing maps are proposed as a tool for dialect research, and a self-organizing map is implemented for the purposes of testing this. Several areas for further research are identified, including how dialects are stored in the brain, more detailed descriptions of how dialects vary, including contextual effects, and more sophisticated visualization tools. Keywords: dialect, accent, identification, recognition, self-organizing maps, words, lexical sets, clustering.