Sign up for our weekly newsletter!
Ilya is a JSD candidate at UC Berkeley School of Law. His research focuses on natural language processing and machine learning applications that are motivated by both theoretical and practical questions in the legal domain.
Topic: Learning in, from, and with Taxonomies
Abstract: In many common NLP and other machine learning tasks, labels come from taxonomies, that is sets organized into hierarchical tree-like relational structures. Taxonomy formalizes semantic constraints that can be leveraged to introduce structural or "soft" inductive bias in machine learning models. They can also be used to verify machine intelligence by assessing certain common-sense semantic relations, beyond commonly used analogy relations (as described in Word2Vec studies, Mikolov et al., 2013). This talk will review definitions of taxonomies, highlight works on graph based and on representation (embedding) based approaches to taxonomy learning, and touch upon several applications to biomedical domain.
Presenter: Dima Lituiev, PhD, Sr Machine Learning Scientist @ Bakar Computational Health Sciences Institute, UCSF
