Monday December 2, 2013 at 16h30 in Celestijnenlaan 200A (auditorium 00.225)
Learning a taxonomy of domain-specific concepts from text
by Daniele Alfarone (PhD student DTAI)
The long-standing goal of deep text understanding requires the availability of taxonomies that organize domain concepts into a hierarchy. Hand-crafted taxonomies necessarily lack coverage and specificity, thus there is a need for learning taxonomies from text. Nevertheless, natural language is imprecise, ambiguous, and contains an impressive amount of implicit knowledge, making this problem very complex.
In this talk, I will present an approach that attempts to tackle these challenges to automatically learn a taxonomy from a domain-specific corpus, without relying on any background knowledge. I will present experiments on several domains, that show how the proposed algorithm overcomes some limitations of existing systems.