Browse Theses

Two issues of database query processing are addressed in this thesis: optimization and representation. Optimization is concerned with query processing techniques to quickly obtain answers from the database; and representation is concerned with query and answer representation between the users and the database systems.

Conventional query processing takes a domain-independent approach to determine the optimal access plan for retrieving the answer. Semantic query optimization (SQO) uses knowledge and semantic reasoning to transform queries into more efficient representations for processing. Integrity constraints (IC) have been used as semantic knowledge for SQO. However, since the purpose of IC is to ensure database integrity, these constraints are often specified in a very general way. As a result, they are of limited value for SQO. Furthermore, acquiring the set of integrity constraints is also a problem. The first part of the thesis presents a model-based learning technique, based on a Knowledge-based Entity-Relationship (KER) model and rule induction techniques, that learns a set of If-then rules from the database contents. Our experimental results show that not all the semantic knowledge is useful for query improvements. The second part of the thesis presents a database restructuring technique, based on type hierarchy and induced rules, which restructures databases to provide a more effective environment for SQO.

Conventional query answers usually are in the form of listing all the instances that satisfy the query. An intensional answer provides characteristics that characterize the extensional answers which gives a summarized description about the answers. The third part of the thesis presents an approach that uses induced knowledge and type inference to derive intensional answers. In a conventional query processing environment, queries have to be rigidly specified and only data satisfy the query will be considered as answers. Cooperative query processing allows vague queries to be specified and approximates data to be provided as answers when the exact answer is not available. The last part of this thesis presents an approach of using type abstraction hierarchy to provide cooperative answers.