NLTP Content Meetings – September 9 – Lexical Semantic Change Detection for Ancient Greek by Silvia Stopponi

We’re opening this academic year’s content meetings on September 9 with a talk by Silvia Stopponi, PhD candidate at the Center for Language and Cognition, University of Groningen, on Lexical Semantic Change Detection for Ancient Greek. See the details below.

Title: Lexical semantic change detection for Ancient Greek: dataset creation and evaluation of a word-embedding-based technique

Speaker: Silvia Stopponi, PhD student, Center for Language and Cognition, University of Groningen

Abstract: Lexical semantic change is a relevant phenomenon from both a linguistic and a cultural point of view, given that the changing usage of words could signal a cultural change. However, computational methods for semantic change detection have been mainly applied to modern languages (an overview is in Tahmasebi et al. 2021). The situation is different for ancient languages, such as Ancient Greek, due to limited corpus size and to lack of native speakers. Most studies about semantic change in ancient languages have indeed been carried out with the ‘philological method’, the manual examination of the occurrences of target words in context.
I assess the viability of semantic change detection for Ancient Greek by using computational language models and two measures of change. In particular, I present a new benchmark for the evaluation of lexical semantic change detection in Ancient Greek and how we used it to assess the validity of two metrics of change, the Vector Coherence and the J, on diachronic embedding models. I also discuss the qualitative evaluation of the detected cases of semantic change by using the tested method.

Reference: Stopponi, S., Nissim, M., Peels-Matthey, S. Forthcoming. Lexical semantic change detection for Ancient Greek: dataset creation and evaluation of a word-embedding-based technique. Computational Linguistics in the Netherlands Journal 14. https://www.clinjournal.org/clinj/article/view/209/218