| |
Abstract:
We present the design, implementation and simulation results
of a psycholinguistic model of human syntactic processing that
meets the major empirical criteria. The parser, a major revision of
a model published in 1989, operates in conjunction with a
lexicalist grammar and is driven by syntactic information
associated with heads of phrases (strictly head-driven parsing).
The most important innovation is a structure optimization mechanism
based on competition by lateral inhibition ('competitive
inhibition'). Input words activate lexical frames (elementary trees
anchored to input words) in the mental lexicon, and a network of
candidate 'unification links' is set up between frame nodes. These
links represent tentative attachments that are graded rather than
all-or-none. Candidate links that, due to grammatical or 'treehood'
constraints, are incompatible, compete for inclusion in the final
syntactic tree by sending each other inhibitory signals that reduce
the competitor's attachment strength. The outcome of these local
and simultaneous competitions is controlled by dynamic parameters,
in particular by the Entry Activation and the Activation Decay rate
of syntactic nodes, and by the Strength and Strength Build-up rate
of unification links. In case of a successful parse, a single
syntactic tree is returned that covers the whole input string and
consists of lexical frames connected by winning unification links.
Simulations will be reported of the essential psycholinguistic
parsing phenomena in both normal and aphasic speakers of
English:
* various effects of linguistic
complexity (single versus double, center- versus right-hand
embeddings of relative clauses, the difference between relative
clauses with subject (easier) and object (harder) extraction, the
contrast between a complement clause embedded within a relative
clause (harder) versus a relative clause embedded within a
complement clause (easier);
* effects of local and global, lexical and syntactic ambiguity
(including effects of recency and constituent length, and the
contrast between local ambiguities that are easy to resolve versus
ones that lead to serious garden-path effects);
* context effects of a semantic (plausibility) or syntactic nature
(structure priming); and
* effects of agrammatism on parsing performance (in particular the
performance of various groups of aphasic patients on several
sentence types; data from Caplan & Hildebrandt, 1988).
To our knowledge, the parsing model to be presented is the first
one capable of simulating parsing performance not only by normal
language users but also by agrammatic patients on a broad range of
syntactic structures.
|
