Dr Guy Emerson

My largest body of published work is on computational semantics.  How do people understand language, and how do people learn to do that?  I approach this with a foot in two worlds: the logical world of formal semantics, and the data-driven world of distributional semantics.  The aim of formal semantics is to develop mathematical models of meaning, with a particular focus on semantic composition and logical inference.  The aim of distributional semantics is to develop computational models of meaning, using algorithms that can be run on a corpus of text.  Combining the two opens up new opportunities.  From the computational perspective, formal semantic structure enables a model to learn and generalise more effectively.  From the formal perspective, a computational model allows us to tackle research questions that would be impossible to handle with pen and paper.

For a gentle introduction to my work, see the following one-hour talk I gave at the ILFC Seminar jointly organised by Université Paris Cité and Université du Québec à Montréal: "Learning meaning in a logically structured model: An introduction to Functional Distributional Semantics"

A constant challenge in the above line of research has been the tension between linguistic expressiveness and computational tractability.  This has pushed me to reconsider basic tenets of probabilistic modelling, and to wonder what a better approach might look like.  My current research focus is understanding fundamental computational constraints on inference processes (including any inference processes in the brain).  The aim is to explain how apparent inconsistencies in human behaviour might naturally arise as a result of computationally constrained minds interacting with a computationally demanding world.

I have more general research interests beyond the above topics, including: machine learning (how can models work with structure?), philosophy of language (what does it mean to know a language?), morphosyntax (what are the components of language?), and NLP for low-resource languages (how can we make sure NLP works for everyone?).  For an overview of how I see the connection between machine learning and language, see this article on "Language and AI".

I am also keen to support researchers in the humanities and social sciences who would like to use machine learning to further their work.  If you would like to discuss any ideas (whether you're at an exploratory stage, or looking for technical feedback), please don't hesitate to get in touch!

For current Cambridge students: I supervise MPhil / Part III dissertations and also Part II dissertations.  See here for previous project suggestions, but feel free to get in touch to discuss any ideas that broadly fit with my research interests.

For prospective PhD students: I am looking for students to work with me on topics connected to Functional Distributional Semantics (see "Publications" tab above).  I would also be interested in supervising topics where there is a clear linguistic research question, and computational modelling is important for answering that question (see "Research" tab for examples).  Feel free to get in touch to discuss your ideas!  To show that you've read this page, please include "hypothetical rhubarb" in the subject line of your email.