I employ experimental, computational, and theoretical approaches in my research. On the empirical side, I explore the brain basis of semantic processing (particularly of argument and event structure) using magnetoencephalography (MEG). On the computational side, I have created a corpus of natural language inference, as well as utilizing semantic vector representations to investigate questions about argument structure. I have also trained an othographic statistical classifier. On the theoretical side, I investigate issues relevant to the syntax-semantics interface, e.g., the connection between argument and event structure in English and Mandarin Chinese, the nature of the connection between number, definiteness, and kind reference.
My most recent experimental work uses MEG to investigate the neurobiological basis of relationality, or the ability of concepts to establish relationships between multiple entities. Prior hemodynamic research points to the left Angular Gyrus as a region that is activated for verbal argument structure (e.g., Thompson et al 2007). However, is it the eventivity of verbs, or the number of arguments (or both) that is driving activity in this region? My research finds that it is relationality, not eventivity that drives left AG (Williams et al, 2017). Furthermore, I investigate a possible connection between relationality and notional plurality, which would postulate that the left AG processes multiples, either of arguments or of discourse entities. We find that, although we can imagine a connection between them, relationality and plurality are likely to be distinct mental processes. Finally, I investigate the extent to which relationality can drive early, visual effects. By implementing an orthographic relationality classifier that takes orthographic n-grams as features, I determine that logistic regression classifiers can bin words as relational or not better than baselines. Then, I determine that there are lower level visual effects (around the M100) of relationality, that suggested there might be a connection between form and meaning in that differs for relational and non-relational words. These projects were advised by Liina Pylkkänen (chair), Alec Marantz, Sam Bowman, Chris Barker, and Andrei Cimpian.
I work on Natural Language Understanding with Sam Bowman (NYU Linguistics/CDS, ML2). We have created a manually-annotated corpus for natural language inference (MultiNLI; the corpus is archived here, the project page is here, and a version write-up is available here), and for use as a resource for developing deep machine learning models that do sentence encoding. This corpus was part of shared task at the RepEval 2017 Workshop at EMNLP in Copenhagen. We also have an (accepted) paper with Andrew Drozdov (eBay) on evaluating latent parse trees that were arrived at by end-to-end sentence encoders learning from our MultiNLI corpus (see arxiv here).
I have an ongoing theoretical semantic project (under review as a volume chapter) that investigates the effect of lexical semantic “constants” on syntactic realizations. It investigates English Weak Definites and Bare Singulars by taking their lack of “uniqueness” to stem from the interaction between definiteness and number within the noun phrase. I put forth a shared analysis of “weakness” that utilizes semantic root ambiguity (Levinson 2014, Rappaport-Hovav & Levin 1998). This account has syntactic consequences that resonate with recent cross-linguistic morphosyntactic accounts of number phenomena that argue NumP is the host of number interpretation and marking (Ritter 1991, 1992, 1995) in languages like Amharic, (Kramer 2009), Halkomelem Salish (Wiltschko 2008), and Haitian Creole (Déprez 2005).
I also had a syntax-semantics interface project on the Mandarin Chinese locative preposition zai when it is used to construct progressive aspect. I explored connections between syntactic argument structure and temporal event structure, while simultaneously presenting novel arguments that this element is prepositional in every case. This qualifying paper project was advised by Stephanie Harves (chair), Alec Marantz and Richard Kayne. Another theoretical project I have been pursuing concerns the interaction between definiteness and number interpretation in American English. Recent developments in this project involve work on PP- and object-shift with Haoze Li, resulting in a poster for LSA 2017.