Lexical signs: associate phonological information with semantics
Grammar rules (also signs): in addition to phonology and
semantics, there is constituent structure
Partial information (feature) structures
Data structures specifying values for attributes
Express constraints on collocations of phonological structure,
syntactic structure, semantic content, and contextual factors
Recursive embedding: the value for an attribute may be another
information structure
Structure-sharing: one and the same structure may occur as value
of distinct attributes in larger structure
Notation: attribute-value matrices and directed acyclic
graphs
Used to represent syntactic structure (including constituency)
and semantic structure
How Linguistic Information is Used
Unification: from a set of compatible structures, yields
a structure containing all of the information present in the set, but
nothing else
In all kinds of language use (analysis, production, translation,
language games, etc.), specification of a token linguistic object
comes about in a cumulative (monotonic) fashion via interaction
of constraints from several sources
Knowledge itself is declarative: characterizes what constraints
apply independently of what order they apply in
Knowledge is reversible: neutral with respect to analysis
vs. production
Linguistic constraints can be interleaved with constraints from
other domains in the course of language use