Abstract "The Limitations of Modeling Finite State Grammars with Simple Recurrent Networks"

In his book `Mechanisms of Implicit Learning' (1993) Axel Cleeremans describes how finite state grammars can be modeled successfully with connectionist networks namely with Simple Recurrent Networks (SRNs) developed by Jeffrey Elman. However, SRNs cannot be used for modeling arbitrary finite state grammars. In this paper I describe the limitations of this approach.


[pdf] [ps] [ps.gz] Publications overview
Last update: September 04, 2000. erikt@stp.ling.uu.se