Research & Output

I consider myself a descriptive grammarian who wants to understand his data, and a computational linguist who loves puzzle solving. To that end, I use a mix of theoretical and computational tools to understand different areas of morphophonology. I have worked on different areas of morphophonology, including:

  • The morphophonology of Armenian
  • Computational morphophonology of…
    • Reduplication
    • Templatic Morphology (Arabic)
  • Acoustic documentation of phonological prominence

You can find out more below. For my dissertation, I am working on the computational morphophonology of Armenian under the supervision of Dr. Jeff Heinz and Dr. Christina Bethin. Armenian is understudied but has morphophonological processes that are sensitive to multiple linguistic factors. I discover these factors in my thesis and I work out how computationally powerful morphophonology needs to be.

Morphophonology of Armenian

Armenian is an understudied Indo-European language with an odd mix of typological properties. I focus on the various ways morphology and phonology interact within the word. I mostly focus on stress and stress-based phonological processes in Eastern and Western Armenian, with some excursions into neighboring endangered dialects.

Over the years, I have gathered data and developed theoretical models to capture:

  1. vowel reduction within sublexical prosodic constituents
  2. vowel reduction in cyclic word-formation and derivation
  3. vowel reduction and subregularities in syllable structure
  4. bracketing paradoxes in compounds
  5. phonologically and syntactically conditioned affix mobility (with Nikita Bezrukov)

To learn more these points, check out:

  • [1] CLS53 paper and slides
  • [2]  PLC42 paper and slides (includes data presented at ConCALL-3)
  • [3-4] TBA
  • [5] TU+ slides (incorporates OV-IS-2018) and PLC43 slides (more discussion of theoretical machinery)

Computational morphophonology

Language is a pattern, and patterns must be computable. But, different patterns need different power.

Computational morphophonology of reduplication

Reduplication is a typologically common yet computational difficult process to model. Under the supervision of Dr. Jeffrey Heinz, I have explored the use of uncommonly used finite-state technology (2-way finite-state transducers) in handling the high computational complexity of reduplication. Using 2-way FSTs, we have:

  1. looked at its ability to generate the typology of reduplication
  2. discovered subclasses of 2-way FSTs that match the linguistic typology
  3. determined the learnability of these subclasses.
  4. shown how they are an insightful computational model of reduplication
  5. developed the RedTyp database on reduplication using 2-way FSTs

RedTyp can be accessed on our GitHub. To learn more about these points, check out:

Computational morphophonology of templatic morphology

I’ve recently looked at the computational power needed to model templatic morphology or root-and-pattern morphology in Semitic. In a pilot study with Jonathan Rawski and Sedigheh Moradi, we are discovering how the subregular hierarchy for single-tape finite-state transducers can be extended to multi-tape transducers.

Acoustic documentation of phonological prominence

As a (long-distance) member of Dr. Irene Vogel‘s Prosodic Typologies lab, I have helped the lab carry out a cross-linguistic acoustic documentation of phonological prominence. Because of methodological difficulties in making comparative analyses of acoustic work on stress, the lab’s goal is to use systematically use the same rigorous experimentation methodology in documenting the word-level and sentence-level prominence of both well-studied and under-studied languages.

Under the supervision of Dr. Vogel, I have carried out an acoustic documentation of prominence in Armenian. The results were included in our Interspeech 2017 proceeding.

I am currently working on expanding the project to include Uzbek and a couple of other languages.