Distributional Semantic Models
Stephen Pulman · Oxford University Press eBooks · 2013
A recent preoccupation in computational linguistics has been the attempt to combine compositional models of formal semantics of the Montagovian type with corpus-based distributional models of word meaning. This chapter describes some of the relevant history and background to this attempt, and surveys some recent approaches to ‘compositional distributional semantics’. It is argued that a failure to distinguish composition proper from one of its frequent side-effects, word-sense disambiguation, means that these approaches fail to demonstrate genuine compositionality. The chapter concludes with some suggestions as to how genuinely compositional distributional semantics might be achieved.