Cross-lingual Lexical Semantic Change in Romance Languages

AbstractWe present a comprehensive quantitative analysis of lexical semantic change in the five main Romance languages (Romanian, Italian, Spanish, French and Portuguese), based on the most exhaustive database of related words in these languages. We include both cognate words and borrowings (for the first time, to our knowledge), and compute semantic shift measures using different static and contextual embedding models, as well as three different corpora. We publish the obtained lists of semantic divergences across all related word pairs, compute global trends in language-level semantic divergence, and provide insights on particular study cases of highly stable and highly divergent words for different language pairs.