I want to build richer models of language. I'm currently advised by Mark Steedman at the University of Edinburgh, where I'm researching novel methods in computational semantics for detection of negation scopes and presupposition. This is useful for understanding processes in the world, for example "being guilty/not guilty" presupposes "being charged".

• Rare Words, Part 3: Recent Methods to Improve Learned Representations
  We'll review recent approaches to word embeddings that focus on improving representations for rare words with sparse contextual information.
• Rare Words, Part 2: Word Vector Semantics with Restricted Vocabulary
  We dive deeper into our previous models to explore semantics of the learned word vectors. What role does vocabulary size play?
• Rare Words, Part 1: Discovering Order in Natural Language
  We'll explore statistical properties of natural language and develop a sentiment analysis model, leveraging what we uncover to drastically reduce the necessary number of parameters.
  Read this on Medium

• How Linguistic Metaphor Scaffolds Reasoning
  Metaphors guide us to think differently about target concepts by borrowing structure from a source concept.
• Semantic Structure and Interpretability of Word Embeddings
  The authors learn a projection matrix from word embeddings to interpretable categories.
• Mental Models and Human Reasoning
  How do humans reason? Rather than perform a routine of logical calculations, perhaps we make simulations colored by our own knowledge.