Automatic language profiling of a dialect speaker: the case of the Timok variety spoken in the village of Berčinovac (Eastern Serbia)

In a previously published paper [Konior et al. 2019], which thematically led up to the present article, we explored the possibility of developing a quantitative tool for assessing the intrasystemic dialectal coherence and the degree of dialectal authenticity (preservation) for a particular variety of Slavic (and more broadly Balkan) dialectal speech. In order to do so, we analysed and manually counted all cases of presence or absence of specific phonemes, direct and indirect object reduplication, ways of expressing peripheral cases meaning, presence of a postpositive article, and some other language features. The data used for that purpose was extracted from “Linguistic Atlas of Eastern Serbia and Western Bulgaria” [SAOSWB]; an idiolect of a native speaker of the Timok dialect spoken in the village of Berčinovac (near the town of Knjaževac in the Zaječar district, Eastern Serbia) was chosen for analysis. Subsequently, the following question arose: how can the use of modern technologies for automatic text processing increase the efficiency of dialectologists’ work, and what technical obstacles must be overcome in this regard? In the article, we present a method of (semi-)automatic analysis of phonetic and morphosyntactic features in a dialect text with the use of morphological annotation (the tagger model is based on the ReLDI tagger [Ljubešić et al. 2016] and user Python scripts). An algorithm searching for some important dialect features is described and exemplified. Trying to imitate and automate historical and structural linguistic analysis, we open a discussion about the advantages and disadvantages of computer analysis of dialect data as compared with the manual analysis. In the future, the automatic method is expected to be helpful in managing larger amounts of dialect data.