Abstract
Combination of Artificial Intelligence (AI) and Connectionist model is very effective to construct an intelligent information processing system. There are several studies based on such a concept. An example, a knowledge base system based on connectionist model has the following features.
(1) Robustness: even when there are some errors in a user's question or in a knowledge base, a near right answer can be obtained.
(2) Context dependecy: retrieval with prediction is possible.
(3) Easy parallel retrieval for multiple concepts.
(4) Easy maintenance of knowledge base: facts are expressed by symbols.
In spite of the above advantages, the conventional system has a great shortcming: it can only make inference of facts.
This paper proposes a connectionist model data base system with a tamplate for association. The proposed system has two features: inference is possible when data is insufficient, and new knowledge can be generated by inference. The proposed system uses a template to create a network for associtive reasoning, and it can be done by interactive activation and competition process. Computer simulation result indicates the effectiveness of the proposed system.
