Oriana Kilbourn-Ceron

Computational linguist working on conversational AI, speech and language data

Portrait of Oriana Kilbourn-Ceron

My work brings together two things: designing conversations between people and machines, and analysing speech and language data with computational and statistical methods.

I joined Desjardins in 2022 as a Conversation Designer, working on a voice-based virtual assistant that pairs conversation design with statistical language models. Before that I was a Postdoctoral Fellow in Linguistics at Northwestern University, working with Matt Goldrick in the SoundLab, after completing my PhD at McGill University. I have taught linguistics at Concordia and McGill, both in Montréal.

My research asks how speakers plan what they are about to say, and how that planning shapes the way the words come out.

What I do

Computational analysis of speech and language

Corpus analysis at scale and automated acoustic measurement, in Python and R. I have modelled pronunciation patterns across a large corpus of spontaneous Japanese speech and worked on an automated method for measuring voice onset time in bilingual speech. I also build tools for handling large volumes of audio data from speech production experiments, including detection of phonetic variants.

See: Japanese vowel devoicing in a speech corpus, Automated acoustic analysis

Experimental design and statistical modelling

Pre-registered speech production experiments, simulation-based power analysis and mixed-effects regression. My data and analysis scripts are openly available.

See: Speech planning experiments, Data and scripts on OSF

Conversation design and language technology

Designing dialogue for a voice-based virtual assistant, where conversation design meets statistical language models. Linguistic theory shapes how I think about what a speech system hears and what it says back.

Selected publications

All publications