Research

I work on unifying logical, computational, cognitive and statistical perspectives on meaning in natural language with Bayesian models of pragmatic reasoning.

I’m interested in scaling these models in computationally tractable ways to handle the complexity of natural language, as well as extending them to figurative language (verbal irony and metaphor) and sociolinguistic indexicality.

Computational Pragmatics

Lost in Machine Translation: A Method to Reduce Meaning Loss (poster)
(NAACL 2019 - Cohn-Gordon and Goodman)

Pragmatically Informative Image Captioning with Character-Level Inference (slides)
(NAACL 2018 - Cohn-Gordon, Goodman and Potts)

An Incremental Iterated Response Model of Pragmatics
(SCiL 2019, ACL Proceedings - Cohn-Gordon, Goodman and Potts)
(Similar work presented at CompPrag 2018 - Cohn-Gordon and Potts)

Communication-based Evaluation for Natural Language Generation
(SCiL 2020, ACL Proceedings - Newman, Cohn-Gordon, and Potts)

Figurative Language

Verbal Irony, Pretense, and the Common Ground
(submitted for publication, 2019 - Cohn-Gordon and Bergen)

Social Meaning

Modeling “Non-literal” Social Meaning with Bayesian Pragmatics (slides)
(Sinn und Bedeutung 2018 - Cohn-Gordon and Qing)

Non-descriptive/use-conditional meaning in Rational Speech-Act models
(Sinn und Bedeutung 2018 - Qing and Cohn-Gordon)

Past Work

Intransitive Object Marking in Amharic (description)
(Presented as a poster at LSA 2017)

Ability Modals (description)

Monads for NL Semantics (draft)

Resultativity in Latin (description)