In the junior research group on ``Computational Modelling of Discourse and Semantics'' we aim to address this task, focussing on the development of integrated models, which bridge the gap between different levels of linguistic structure. One area where two levels of linguistic structure interact is in the recognition of idioms, such as "spill the beans" or "let the cat out of the sack". While idioms are often treated as a phenomenon of lexical semantics, I will show how discourse structure contributes to the interpretation of an idiomatic expression and how this can be exploited to automatically detect idiomatic (i.e. non-literal) language use in a text.