Measuring semantic specificities across corpora: looking for semantic shifts in Quebec English

Résumé En

In this talk we will present a number of works aimed at identifying contact-induced semantic shifts in Quebec English. Our target comprises words whose meaning has been influenced by the contact with French in the specific multilingual context of the French-speaking Canadian province. For example, "deception" is sometimes used by English speakers from Montréal with the meaning of "disappointment" (instead of "dishonesty"), because of its formal similarity with French "déception". We followed different approaches and techniques (based on frequencies and word embeddings) on a large corpus of tweets collected from users in Toronto, Vancouver and Montreal. We will discuss the difficulties encountered when dealing with rare and fine-grained semantic variations across subcorpora.