Abstract:
Forced alignment systems are commonly used to process large amounts of data in socio-phonetic studies. We compare how two of these systems perform against manual segmentation and phonetic labelling in a database of New Zealand English. The results indicate predictable variations in terms of the relative sizes of the vowel spaces, but also suggest the need to be cautious in interpreting small phonetic variations, as these may be the result of the method used to segment and label the data. Index Terms: sound change, forced alignment, artifact, vowel