This paper presents the architecture of a neural network expert system
shell. The system captures every rule as a rudimentary neural network
, which is called a network element (netel). The aim is to preserve th
e semantic structure of the expert system rules, while incorporating t
he learning capability of neural networks into the inferencing mechani
sm. These netal rules are dynamically linked up to form the rule-tree
during the inferencing process, just as a conventional expert system d
oes. The system is also able to adjust its inference strategy accordin
g to different users and situations. A rule editor is provided to enab
le easy maintenance of the netal rules. These components are housed un
der a user-friendly interface. An application expert system for US fut
ure bonds trading is built upon this shell. The connectionist expert s
ystem has demonstrated its strength over the conventional rule-based s
ystem.