Connectionist Parsers
- Connectionist systems are neurally inspired. MLPs,
SOMs, and RBFs are examples of these.
- We're working with models of neurons, so they're neural
models and connectionist models.
- You can learn some things from connectionist systems so:
- Mikkulainen 93 had a recurrent MLP system but these have
problems with longer systems.
- Henderson 94 used SHRUTI to parse. Nice work with
variable binding by synchrony.
- Kempen and Vosse 91 used simulated annealing to resolve
attachment decisions.
- Tabor and Tannenhaus 99 used attractor basins to
resolve decisions.
- The only other neural parser I know of us Knoblauch 2004
and an earlier version of my system.
