A large and growing body of work explores the use of semantic
annotation of business process designs, but these annotations can be
difficult and expensive to acquire. This paper presents a data-driven approach
to mining these annotations (and specifically post-conditions) from
event logs in process execution histories which describe both task execution
events (typically contained in process logs) and state update events
(which we record in effect logs). We present an empirical evaluation,
which suggests that the approach provides generally reliable results.