SN+M: A TABLE-AUGMENTED SEMANTIC NETWORK FOR DATABASE APPLICATIONS (INHERITANCE RELATIONS, SEMANTIC NETWORK PLUS MODEL)
Abstract
Conventional data models have not proven adequately powerful to model a more complex data domain. Focusing on the marriage of technologies from both artificial intelligence (AI) and database (DB) in order to enhance data modeling capability, we have proposed the Semantic Network plus Model (SN('+)M) to tackle the problems caused by using record-based data models. SN('+)M is the result of coupling the semantic network in AI and the relational model in DB such that it possesses network and table capabilities to powerfully and naturally accommodate the features in a complex data domain. An SN('+)M consists of two parts: the semantic network part and the table part. The parts are connected by links showing the semantics of data in tables. The network part is used to enable the database designer to incorporate naturally and directly a large portion of semantics in the entire database. The table part is used to represent tabulated information, for which the record-based table method provides the best model. We also discuss the properties of ISA, AKO and ISPART inheritance relations, which are the backbone of the network part in SN('+)M, based on two types of information--class and object. The properties of these inheritance relations are explored from the from-class-to-class, from-object-to-class and from-object-to-object relations, respectively. Finally, we have used a CAD/CAM application domain to demonstrate how SN('+)M can be used to powerfully and naturally model a complex data domain consisting of different data types: formatted and unformatted forms, static and dynamic status, record and event information. Much of the library, design and manufacturing information, such as CSG tree representation in geometrical modeling, bill of materials in parts assembly, assembly and machining events and precedence relations among assembly and machining operations, can be completely and efficiently modeled by SN('+)M with simple and direct algorithms.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.