Marco Baroni & Georgiana Dinu

Language & Computation

Week Two - 14.00-15.30 - Level: A

Room: 7E02


Distributional semantic models, that represent the meaning of words in terms of their patterns of co-occurrence in corpora, have been very successful at modeling various aspects of lexical semantics, such as detecting synonymy or predicting the selectional preferences of verbs. In recent years, there has been much interest in extending these models with a compositional component, to derive distributional representations of phrase and sentence meaning. In this course, after providing the relevant background in both distributional and compositional semantics, we will introduce the main current approaches to compositional distributional semantics both theoretically and hands-on. We will present the tasks that have been tackled using these methods and discuss future directions as well as broader implications for theoretical and computational linguistics and cognitive science.