Ewan Dunbar

Research

My research is in phonology. Phonology studies those components of human cognition that relate directly to recognizing speech and pronouncing it, in particular the higher level systems we call phonological grammar. My research uses computational tools (especially Bayesian statistics) to help us learn about those systems. I have looked especially at the interaction between how humans learn low-level speech sound categories and how they might learn more abstract facts about their phonology, which has implications for phonological theory.

Papers and Talks

For a slightly longer list, download my CV.

A single stage approach to learning phonological categories: Insights from Inuktitut. With Brian Dillon and Bill Idsardi. Under review.

Short description of infinite Bayesian mixture of linear models available here. Code available below.

Review of Daniel Silverman, ``A Critical Introduction to Phonology''. With Bill Idsardi. 2010. In Phonology 27:325-331.

A single-stage computational model of phoneme category acquisition. With Brian Dillon and Bill Idsardi. Poster presented at Computational Modeling of Sound Pattern Acquisition Workshop, University of Alberta, 2010.

Seeing Through the Surface: A Model for Direct Acquisition of Phoneme Categories. With Brian Dillon and Bill Idsardi. Poster presented at Boston University Conference on Language Development, 2009.

Pitfalls of Distributional Allophone Learning. (PDF with notes and extra slides. Original OpenOffice.org Presentation available on request.) Presented February 28, 2009. Montreal-Ottawa-Toronto Phonology Workshop (MOT) 2009.

The Acquisition of Morphophonology Under a Derivational Theory: A Basic Framework and Simulation Results. 2008. Forum paper (MA thesis), University of Toronto.

Developing Intermediate Language Learning Materials: A Labrador Inuttitut Story Database. With Joan Dicker and Alana Johns. Presented May 3, 2008, Stabilizing Indigenous Languages Symposium (SILS) 15, Northern Arizona University. In Indigenous Language Revitalization: Encouragement, Guidance and Lessons Learned. (Proceedings of SILS 14 and 15.)

Where Will the Stories Go? With Alana Johns. March 18, 2008. Talk, Inuktitut Linguistics Workshop, University of Toronto.

Teaching

Discrete mathematical models of mental grammar (LING499B). Upper-year seminar, Spring 2012.

Basic stats and R crash course. Originally part of UMD IGERT Winter Storm 2011. Co-taught with Alex Drummond.

Basic experimental data analysis in six hours. Originally given at Stony Brook University, October 1-2 2010.

Code

dmix.tar.gz. R package containing code for Dillon, Dunbar, and Idsardi (under review). Email to obtain.

Useful Links

Customizable Presentation of ELAN Documents (CuPED). ELAN is already a great program for doing language transcripts. Now these folks have developed a system that can export to arbitrary templates, most crucially a Flash/JS-ified web page. Fantastic for archiving, and customizable. Related to our Inuttitut Story Database (see the paper above, contact me for more details); fills a different niche, though. Check it out.

RStudio. R is a fantastic statistical program, but the GUIs that ship with R (like R.app on the Mac) do little but eat extra system resources. RStudio is almost a full IDE: it has a much better editor, it allows you to visualize common data structures, organizes your plots into a nice stack of tabs, and takes care of all the little things that sometimes get in the way (like re-source your files manually each time you save them when working in interactive mode). It's missing a debugger (but R does have one) and it's a bit sluggish (it uses Java+Google Web Toolkit), but, for whatever it's worth, I would never consider going back to vim for this purpose.