Lexical semantics: bridging the gap between semantic theory and computational simulations

Corpus-based distributional semantic representations (like LSA or HAL) capture many aspects of word meaning. However, no single computational approach meets all requirements posed by formal semantics or linguistic theory, nor do they cater for all aspects of meaning that are important to philosophers or cognitive scientists. We invite developers to test their model on a variety of tasks inspired by the theoretical and cognitive literature (e.g., distinguishing natural kinds from artifacts, extracting salient concept properties, qualia identification, argument structure). Test sets will be kept small to facilitate qualitative analysis. The focus is not on competition, but on understanding how different models highlight different semantic aspects, and how far we are from integrated models of all such aspects. Theoretical and experimental papers related to the task datasets and simulation results are also invited. The workshop will bring together research communities that work on very similar themes from sometimes very different perspectives.