SpeakerEkaterina Shutova
DateDec 04, 2015
TitleStatistical modelling of metaphor
AbstractBesides making our thoughts more vivid and fillingour communication with richer imagery, metaphor plays afundamental structuralrole in our cognition, helping us organise and project knowledge.For example, whenwe say “a well-oiled political machine”, we view the concept ofpoliticalsystem in terms of a mechanism and transfer inferences from thedomain ofmechanisms onto our reasoning about political processes. Highlyfrequent intext, metaphorical language represents a significant challenge fornaturallanguage processing (NLP) systems; and large-scale, robust andaccuratemetaphor processing tools are needed to improve the overallquality of semanticinterpretation in today’s language technology. In this talk I willintroducestatistical models of metaphor identification and interpretationand discusshow statistical techniques can be applied to identify patterns ofthe use ofmetaphor in linguistic data and to generalise its higher-levelmechanisms fromtext.
BioEkaterinaShutova is a Leverhulme Early CareerFellow at the University of Cambridge Computer Laboratory. Herresearch is inthe area of natural language processing with a specific focus oncomputationalsemantics and figurative language processing using statisticallearning.Previously, she worked at the International Computer ScienceInstitute and theInstitute for Cognitive and Brain Sciences at the University ofCalifornia,Berkeley and the Department of Theoretical and Applied Linguisticsat the Universityof Cambridge. Ekaterina received her PhD in Computer Science fromthe Universityof Cambridge in 2011 and her doctoral dissertation concernedcomputationalmodelling of figurative language.