A Framework for Network Modeling

The paper is a respond to the recently growing quest for hybrid approaches trying to integrate various AI paradigms in order to overcome their limitations. It presents a network modeling environment, called Net-clause language (NCL). NCL ooers distributed computational environment using data-driven control and spreading activation. Though it falls basically in the symbolic paradigm, NCL exhibits some connectionist features. However, in contrast to the majority of connectionist/symbolic hybrids, NCL is not an "imple-mentationalist" approach. Rather it uses some connectionist ideas (such as thresholds, excitatory and inhibitory connections) to organize symbol representation and processing in a more exible way. Various aspects and applications of NCL have been discussed so far elsewhere. The present paper is an attempt to put them together by discussing an approach to natural language parsing and disambiguation.