At this week’s MCQLL meeting, Gaurav Kamath will be presenting “Testing the Apparent Time Construct in Embedding Space”. Abstract below.
We will be meeting this Tuesday October 31st at 3:00PM. Meetings are held both in person in room 117 of the McGill Linguistics department and on zoom.
Abstract: Linguists are often interested in studying how language changes over generations. But studying these patterns in real-time is often unfeasible, because it would either require tracking one population group over several years, or somehow going back in time to observe old speech patterns. Sociolinguistic analyses of language change over time therefore tend to rely on what is known as the apparent time construct: a methodological construct that involves taking the current speech patterns of older speakers (e.g. a 60 year old in 2023) as representative of the speech patterns of younger speakers at a prior timepoint (e.g. a 20 year old in 1983). While the validity of this construct has been tested in terms of phonetics/phonology (e.g. Bailey et al. 1991, Freuhwald 2017), it has so far not been properly tested in terms of meaning change. On the other hand, diachronic word embeddings have been shown to capture certain aspects of semantic change (Hamilton et al. 2016). In this talk, I present early-stage plans and efforts towards testing the apparent time construct vis-à-vis semantic change, by using word embeddings from models trained on historical corpora.