A Computational Analysis of Gabon Varieties

The linguistic situation in Gabon is highly complex as the various varieties form long chains, and multilingualism is common. Most previous classifications of Gabon varieties have used lexical models, a notable exception being Nurse & Philippson (2003), which involves phonological and morphological features. This paper presents a phonetic analysis of Gabon varieties. It applies Levenshtein analysis to obtain pronunciation distances, which are in turn analyzed using bootstrapped clustering to identify groups, including an estimation of the robustness of the clusters. On the basis of clustering, we obtain a graded map of varieties. The results indicate that techniques developed and proven on European languages are still useful when applied to Bantu in spite of its different structure and its nomadic speakers.