I recieved my PhD in Linguistics from the University of Southern California in 2021. My dissertation is entitled “Soft Biases in Phonology: Learnability meets grammar”. My work focuses on using computational and mathematical methods to understand how the process of learning phonological structure influences the likelihood of certain structures appearing in the languages of the world. Particularly, my work is on the ways learning can gradiently affect typology, and on the development of algorithms to learn underlying representations. From Winter 2021 to Summer 2022, I was a Lecturer in the department of Linguistics at the University of Michigan.

My Research Interests

• Phonology
  • Phonological learning

  • Learning’s effect on soft typology.

    <summary> Expand</summary> I use simulations to identify which grammars in a typology would be most consistently learned accurately across generations, and to understand the subtle interactions of learning and grammar. (This is the topic of my dissertation; See my poster at AMP 2017, and my LSA 2018 talk on this. Other chapters of my dissertation explore how lexical frequency of different forms affect learnability (Workshop on Analyzing Typological Structure, AMP 2018), as well as whether substantive biases can be restricted to the channel (Workshop on the Emergence of Universals).

  • Abstractness in Underlying Representations.
    <summary> Expand</summary> Analysts have long proposed underlying representations that are not immediately apparent from the surface forms of a morpheme (in order to handle seemingly exceptional processes, etc.) What I find, is that several of the analytical influences that have been used to motivate use of abstract URs (i.e. avoiding accidental gaps) are emergent results in a MaxEnt learner. (See How Abstract is More Abstract, Phonology 34.2 2017)
  • What can phonological learning theory tell us about child language acquisition?
  • Typological differences between Optimality Theory and Harmonic Grammar.
    <summary> Expand</summary> Within the last decade or so, many phonologists have begun recognizing some of the merits of using weighted constraints rather than ranked constraints. With this conversion, many of the constraints used in OT can have surprising interactions that can be troublesome or beneficial. (See [Sonority Based Stress in Harmonic Grammar] and Harmony in Harmonic Grammar by Reevaluating Faithfulness)
  • Mathematical approaches to phonology
    <summary> Expand</summary> With Caitlin Smith, I have been investigating how the computational complexity of feature spreading patterns; showing the existence of a sub-regular class of mappings that includes all attested feature spreading patterns, but not some unattested but regular mappings, like sour grapes. See our SCIL 2019 extended abstract.
• Ethics of Large Language Models and AI approaches to language.
• The Klamath language, a recently extinct language from Southern Oregon
• Semantics
  • Nouns modifying adjectives in English Hillary Clinton is rich, but not Mitt Romney rich. (See Nouns Attributively Modifying Adjectives in English and Hilary Clinton is not Mitt Romney rich)