In this course, I explore the interplay between theory and
experimentation in linguistics. Specifically, I present the current
theory of argument selection within Generative Lexicon Theory
(Pustejovsky, 2007), where "compositionality as function application"
is enriched with operations of coercion and accommodation. I then
examine how these theoretical statements and generalizations of
language actually fair when confronted with linguistic data from very large
corpora. Some observations survive while others fail. Finally, the methodology
of Model-Annotate-Train-Test is presented as important tonic
to armchair semantic analysis.