Subsymbolic Parsing of Embedded Structures
Short Description
The talk presents a distributed neural architecture SPEC which has learning
capability for parsing sentences with relative clause. The SPEC architecture
was mostly motivated from the artificial intelligence point of view by the
desire to build a system that
- would be able to process nontrivial input like symbolic systems, and
- makes use of the unique properties of distributed neural networks such as
learning from examples, spontaneous generalization, robustness, context
sensitivity, and integrating statistical evidence.
References
-
Subsymbolic Parsing of Embedded Structures,
Miikkulainen, R.,
in "Computational Architectures Integrating Neural and Symbolic Processes :
A Perspective on the State of Art". Sun, R. and Bookman, L. (Ed.), Boston: Kluwer Academic Publishers, 1994