A warm welcome to Daniel Edmiston, who is a visiting PhD student this term. His (self-)introduction follows:
“I’m a PhD Candidate at the University of Chicago in the linguistics department. Broadly, I work on computational linguistics and natural language processing. Narrowly, I’m interested in evaluating continuous representations of linguistic units, such as those produced by neural networks. My dissertation focuses on identifying morphological information contained in continuous word vectors, using techniques from unsupervised learning (addressing issues like intrinsic dimensionality) and computational topology (such as persistent homology, seeking to identify the large-scale structure of word-embedding models). I am tangentially interested in designing and evaluating linguistically motivated neural network models, specifically those that model compositionality by incorporating the notion of type.”