About me

I’m a quantitative historical linguist. I investigate how languages change over deep time, and what their evolution can tell us about culture and human cognition. I’m particularly drawn to Sahul (Australia and New Guinea), which formed a single landmass until only around 10,000 years ago — relatively recently in the 60,000+ year timeline of human history in the area. My work focuses mainly on the Indigenous languages of Australia, with extensions north to New Guinea, Tibeto-Burman, and worldwide typological samples. My toolkit is primarily Bayesian and phylogenetic: tree inference, ancestral state reconstruction, and phylogeographic regression. Two broad questions drive my current work. How do conceptual categories like noun classification, case, and number evolve across thousands of years? And can we distinguish what languages inherited from common ancestors from what they borrowed, adapted, or invented on their own?

A more recent thread of my work sits between linguistics and AI. Much of what’s known about the world’s linguistic diversity is locked up in descriptive grammars, often printed decades ago and often only available as scanned PDFs. Plain-text OCR is tractable, but getting at the structured content that linguists care about (paradigm tables, interlinear glosses, phonological rules) is much harder. Extracting structured data from grammars is a problem I’ve been chipping away at with collaborators since 2016; LLMs open up what wasn’t previously possible here, but getting them to reliably handle idiosyncratic linguistic notation takes careful design and extensive validation. A recent prototype, benchmarked against manual annotation on an Australian grammar sample, hit 99% precision and 99% recall.

I’m currently a Visiting Scholar at the Surrey Morphology Group, University of Surrey. Before that, I was a Lecturer in Linguistics at the University of Newcastle, Australia (2022–2025), and before that a Postdoctoral Researcher at the CNRS Dynamics of Language lab in Lyon. I did my PhD at the University of Queensland. I’m happy to hear from anyone interested in collaboration, consulting, or just a conversation about language, computation, or messy real-world data — jayden {at} macklin-cordes.com.