Adina Williams

image-001

You can view my CV here (updated 12/18).

I am a postdoctoral researcher in linguistics at FAIR NYC (as of Oct. 2018).

As a highly interdisciplinary researcher, I specialize in semantics, syntax, and their interface, with neurolinguistics, computational linguistics, and formal linguistic theory as my main methodologies. Theoretical linguistic insights form the bases for my hypotheses about grammatical representations in the mind and computer. I defended my PhD in Linguistics at New York University in Apr. 2018, and it conferred in September 2018; it is  Representing Relationality: MEG Studies of Argument Structure, and includes 3 MEG experiments, and a chapter on classifying words as transitive-intransitive using only their orthographic forms.

I recently released the Multi-genre Natural Language Inference (MultiNLI) corpus, which is a crowd-sourced collection of sentence pairs (433k) annotated with textual entailment information for use in NLP and Machine Learning applications. It was the basis of shared task associated with the RepEval 2017 Workshop at EMNLP 2017 (Copenhagen); training and test data is available in full on Kaggle as part of an indefinitely open evaluation (for both matched and mismatched subsets). The manuscript describing the data and giving some baselines is available here; a manuscript detailing the results of the shared task is available here. We used MultiNLI as part of some exciting new work on evaluating latent parse trees that were arrived at by end-to-end deep architectures doing Natural Language Inference without reference to explicit parse trees at training (here). Both of these works were presented at NAACL 2018 in New Orleans, LA. MultiNLI is also the basis for XNLI, which extends the development and test sets of MultiNLI to 15 languages, and enables evaluation of cross-lingual understanding and transfer. This work was presented at EMNLP 2018, in Brussels, Belgium.

Interests

Broadly:

  • Topic: Syntax-Semantics Interface
  • Methodologies: Computational & Experimental Approaches to Linguistics

Narrowly:

  • Brain Basis of Syntactic and Semantic Processing
    • Argument Structure and Event Structure
    • Syntactic Category
    • Representations of Number
  • Natural Language Understanding 
    • Universal Sentence Representations
    • Natural Language Inference, Crowd-sourcing for NLP
    • Evaluating syntactic representations induced from TreeRNNs that perform semantic tasks
  • Semantics of Inflectional Morphology
    • Number Interpretation, Marking, and Countability
    • Prepositional and Verbal Aspect
    • Bare Singulars and Weak Definites
  • Syntax & Semantics of Mandarin Chinese
    • Morpho-semantics of aspect (micro-parametric syntactic variation)
    • Classifiers and countability