Syntactic theories relate sentence structure to the details of morphemes, i
nflections, and word order. Semantic theories relate sentences to the detai
ls of formal logic and model theory. But many of the most successful progra
ms for information extraction (IE) are based on domain-dependent templates
that ignore the details at the center of attention of the major theories of
syntax and semantics. This paper shows that it is possible to find a more
primitive set of operations, called the canonical formation rules, which un
derlie both the template-filling operations of IE and the more formal opera
tions of parsers and theorem provers. These rules are first stated in terms
of conceptual graphs and then generalized to any knowledge representation,
including predicate calculus, frames, and the IE templates. As a result, t
he template-filling operations of IE become part of a more general set of o
perations that can be used in various combinations to process knowledge of
any kind, including linguistic knowledge, at any level of detail.