Lexical semantic change detection in Latin: a use-case on medical Latin скачать в хорошем качестве

Lexical semantic change detection in Latin: a use-case on medical Latin

Paola Marongiu (University of Neuchâtel) and Barbara McGillivray (KCL) Friday, Jul 7 2023, 17:00-18:00 Lexical semantic change detection in Latin: a use-case on medical Latin Digital Classicist London seminar Institute of Classical Studies, University of London ONLINE ONLY Lexical semantic change (LSC) is the linguistic phenomenon by which words change their meaning over time. Recently, there has been a growing interest in applying computational resources to investigate LSC. Among these resources, diachronic word embeddings have been applied with success to LSC research in various languages, among which Latin (Sprugnoli et al. 2020). In this seminar, I will report on the results of the evaluation of diachronic word embeddings to trace LSC in Latin, with a case-study on medical Latin. The lexicon of technical or specialised languages generally shows examples of LSC (Roelli 2021). The lexicon of medical Latin in particular is rich in polysemous words that have acquired more specialised meanings, developed from more basic or common ones through various types of semantic change (Langslow 2000; Roelli 2021). Examples are forceps (or forfex) from ‘tongs’ to ‘surgical pincers’, mola from ‘mill-stone’ to ‘molar tooth’, causa from ‘cause (of a disease)’ to ‘disease’ (Langslow, 2000: 155;165;182). Moreover, Latin medical texts cover a diachronic span going from the 1st century BCE (Celsus’ De medicina) to the 19th century CE (von Bene’s Elementa medicinae practicae), which allows for a diachronic study of semantic changes on technical terms. After a short introduction on the application of diachronic word embeddings to LSC detection, I will present the results of this case-study on medical Latin. In particular, I will elaborate on the strategies implemented for the evaluation of the word embeddings, based on previous work on Latin (Sprugnoli et al. 2020). This will comprise i) the presentation of the gold standard dataset built for the evaluation by combining lexical and lexicographic resources for Latin; ii) a discussion on the issues of applying diachronic word embeddings to LSC detection in Latin, specifically on medical Latin lexicon, and possible ways for improvement.