Talk: Towards a computer-aided framework for borrowing detection
Gothic loanwords in Hungarian?
How can computers help us to find old loanwords and is it possible that Hungarian exhibits a Gothic loanword stratum? The model that I will propose in my dissertation is capable of extracting sound change laws from an etymological dictionary, reconstructing pseudo-roots of current Hungarian words based on extracted sound changes, and search for potential sound substituted Gothic words within those reconstructed roots.
It has been noted recently that computational historical linguistics is still in its infancy. This statement is even more true for the subfield of computational historical loanword research, since most efforts still go into researching the inherited lexicon. This might well be due to the complexity of the topic. In my project, I propose a new methodology, partly based on Optimality Theory (Prince and Smolensky 1993), whereby I reconstruct pseudo-roots of Hungarian words and check for identical matches within pseudo-soundsubstituted Gothic words. Opposed to state-of the-art methods that calculate the phonetic similarity of semantically identical hand-picked concepts, the algorithm much rather searches for rule-based phonetic matches, and calculates their semantic similarity only afterwards. This model has two main advantages: Firstly, it is not agnostic to semantic change anymore, and secondly, it acknowledges the fact that related words do not necessarily exhibit similar phonetics but rather stick to concrete established sound change rules