Summary of MSc Thesis

Tools for Integrating Conceptual Schemes in a Multi-database Environment using an Object-Oriented Approach.

The integration of heterogeneous pre-existing database systems is important to provide access to data distributed across systems. Some difficulties arise in the integration process due to conflicting values for related data in different database systems.

To alleviate the problems of heterogeneity, each component should support an intermediate schema in a common data model. Such a common data model should have some desired features to overcome many obstacles.

Recently, Object-Orientation technology seems promising to solve such obstacles in schema translation and integration because of the ability of capturing the semantics of conceptual schemes.

In this project, an Object-Oriented Model has been proposed as a Common Data Model to capture the necessary semantics of the relational, hierarchical, and network data model schemes.

Three transformation rules were designed in order to transform relational, hierarchical, and network data model schemes to the object common data model that represent the intermediate schemes.

It was difficult to capture clear conceptual schema from Pre-relational systems such as CODASYL network system and IMS hierarchical system, because most of the semantics are hidden in the application programs. However, such difficulties may be overcome -simply- by user interaction, or by achieving operational mapping that can be supported by object-oriented paradigm.

The produced intermediate schemes will be integrated through various integration rules to produce a Global Conceptual Schema that describes the structure of the whole nodes participating in the Multi-database system. Integration process, in general, consists of three layers:-

1) The lowest layer compares object properties( attributes and methods) of pair of schemes. This layer produces different weights that represent the number of equivalent properties between the two objects.

2)The Intermediate layer calculates the weighted sum to decide whether two objects are equivalent or not, taking into consideration the user conformation.

3)The upper layer decide whether the pair of schemes are integratable or not, this will be decided according to the entered weights and assertions.

Some operators were built to merge the structure of the equivalent objects and relationships in order to generate a minimal, understandable, and complete Global Conceptual Schemes.

An interactive tool (Tools for Schema Transformation and Integration TOOSTI) was developed using Borland C++ under Microsoft Windows, which supports the various mentioned phases of the schema transformation and integration processes.

The main features of this tool is that it is based on object- oriented methodology. Secondly, The phases of transformation and integration are driven automatically without user intervention, but user may be interrogated in the case of semantic comparison. Third, this tool may be expanded to involve future algorithms and techniques because of the reusability property of Object-Orientation. Forth, object correspondences depend on two measures weighted sum and Assertions. Assertions may be generated due to weighted sum, or user interaction.

Through integration of tens experimental applications using this tool, we found that it is difficult to isolate user interaction during integration process, since we found that two schemes may be integratable, in spite that there are no object correspondences between them. However, we may minimize user interaction by following future work:

1) Object Common Data Model may be augmented by a suitable knowledge-base that support reasoning system of the available assertions.

2) A case-reasoning system can be developed to solve semantic integration problems by utilizing previous solutions.

3) Some of linguistic methods may be adopted to solve serious problems in semantic integration.