Mining the Czech General Service List for Minimal Pairs

Kate Challis

Abstract: The Czech General Service List (CGSL) is a list of 10,650 words with high frequency and wide dispersion across both written and spoken Czech designed to be useful for Czech leaners and teachers. This poster presents the results of an analysis of vowel and consonant contrasts in minimal pairs on the CGSL. First, an R script counted instances of consonant-consonant and vowel-vowel pairs with a Damerau-Levenshtein distance of 1 on the IPA transcriptions of the CGSL words provided by Eufonometr (Plecháč, 2017 [http://versologie.cz]), for example, dělo-dílo. An R script was used to calculate the number of differences in phonetic features; for consonants, these features followed O'Grady et al. (1993) by taking into account manner and place of articulation, and for vowels, these features followed a simplistic binary approach. The graphs on the center of the poster report the counts of minimal pairs on the CGSL for each possible vowel-vowel and consonant-consonant contrast, highlighting one-feature differences in orange and two-feature differences in blue. Previous research has shown that the agreement of articulatory features is considered a measure of learning burden (i.e. functional load) due to perceptual similarity and difficulty of articulation (Tarone, 1978; Celce-Murcia et al., 1996), therefore it is thought that L2 learners should focus greater attention on those contrasts with the greatest number of minimal pairs which are the most phonetically similar. Several examples of minimal pairs from the CGSL are provided on the right side of the poster to illustrate how pronunciation can affect one's ability to be understood in Czech.