Classification of South African languages using text and acoustic based methods: A case of six selected languages

Proceedings of NAACL-HLT 2013, pages 280–287, Atlanta, Georgia, 9–14 June 2013. c⃝2013 Association for Computational Linguistics Classification of South African languages using text and acoustic based methods: A case of six selected languages Peleira Nicholas Zulu University of KwaZulu-Natal Electrical, Electronic and Computer Engineering Durban, 4041, South Africa zulup1@ukzn.ac.za Abstract Language variations are generally known to have a severe impact on the performance of Human Language Technology Systems. In or- der to predict or improve system performance, a thorough investigation into these variations, similarities and dissimilarities, is required. Distance measures have been used in several applications of speech processing to analyze different varying speech attributes. However, not much work has been done on language dis- tance measures, and even less work has been done involving South African languages. This study explores two methods for measuring the linguistic distance of six South African lan- guages. It concerns a text based method, (the Levenshtein Distance), and an acoustic ap- proach using extracted mean pitch values. The Levenshtein distance uses parallel word tran- scriptions from all six languages with as little as 144 words, whereas the pitch method is text-independent and compares mean language pitch differences. Cluster analysis resulting from the distance matrices from both methods correlates closely with human perceptual dis- tances and existing literature about the six lan- guages. 1 Introduction The development of objective metrics to assess the distances between different languages is of great theoretical and practical importance. Currently, subjective measures have generally been employed to assess the degree of similarity or dissimilarity between different languages (Gooskens & Heeringa, 2004; Van-Bezooijen & Heeringa, 2006; Van-Hout & Münstermann, 1981), and those sub- jective decisions are, for example, the basis for classifying separate languages, and certain groups of language variants as dialects of one another. It is well known that languages are complex; they differ in vocabulary, grammar, writing format, syntax and many other characteristics. This presents levels of difficulty in the construction of objective com- parative measures between languages. Even if one intuitively knows, for example, that English is closer to French than it is to Chinese, what are the objective factors that allow one to assess the levels of distance? This bears substantial similarities to the analo- gous questions that have been asked about the rela- tionships between different species in the science of cladistics. As in cladistics, the most satisfactory answer would be a direct measure of the amount of time that has elapsed since the languages’ first split from their most recent common ancestor. Also, as in cladistics, it is hard to measure this from the available evidence, and various approximate measures have to be employed instead. In the bio- logical case, recent decades have seen tremendous improvements in the accuracy of biological meas- urements as it has become possible to measure dif- ferences between DNA sequences. In linguistics, the analogue of DNA measurements is historical information on the evolution of languages, and the more easily measured—though indirect measure- ments (akin to the biological phenotype)—are ei- ther …

Content truncated to 500 words