Date created: | Last Updated:

Category: Project

Description: This is an experiment on prediction of words that have so far not been observed in field work. The basic idea is: with an existing dataset of cognate words, we can use the correspondence pattern detection algorithm by @List2018PREPRINTa to infer for those words in a given language that do not show a reflex in a given cognate set, how they would sound if they were cognate. Since Tim Bodt, who did field work on the Kho-Bwa languages, in fact dit not elicit all words during his initial field work on eight varieties of Kho-Bwa, we can now use the existing dataset to predict how potential cognates would have sounded, and if Tim Bodt goes back to field work in November, he can try to elicit those words and see how well the prediction algorithm works and how well prediction works in historical linguistics in general.