A VISUAL KNOWLEDGE QUERY LANGUAGE FOR THE ENHANCED-ER MODEL

Abstract

One of the major difficulties in the current user-database interface is that the user has to deal with unfamiliar data objects, such as relations and attributes in the database [TeBu86, June86], and has to carefully manipulate many pieces of these data elements in order to specify simple real world concepts. Jones and Shave [JoSh87] state that the required knowledge of the database scheme and the formal syntax of conventional query languages cause problems for naive and infrequent users of the database systems. It is therefore important to relax these requirement, so that the user can manage with partial or even no knowledge of the database structure. This research aims at introducing a new abstraction, knowledge abstraction, to the design of user-database interface. The main thrust of this thesis will be the design and development of a visual query language for user-database interaction. To incorporate the knowledge abstraction to the ANSI/SPARC three-level architecture, this thesis proposes the use of two separate models at the conceptual level, namely the knowledge model and the representational model. The language designed is known as the Visual Knowledge Query Language (VKQL)1. The use of visual approach in VKQL frees the user from the need to memorize the rigid syntax of the query language. Furthermore, empirical studies suggest that visual languages are easy to learn, easy to use, and more appealing than a conventional language [GITa84]. To alleviate the need of the end users to understand large and complex database structures, knowledge abstraction is employed. The main idea of this knowledge abstraction is to separate the knowledge from its storage representation. In other words, with knowledge abstraction, the user specifying a query only need to state the knowledge of the query without bothering about the logical or physical access path. To incorporate knowledge abstraction to the ANSI/SPARC three-level architecture, two different models are proposed at the conceptual level, namely the knowledge model and the representational model. The knowledge model provides a means of describing the database application environment independent of the storage structure whilst the representational model would be used for translating the knowledge scheme into the physical data structure of the internal level of the DBMS. In this thesis, the knowledge model is based on an Enhanced Entity-Relationship (EER) model. The EER model is chosen as the notion of entities and relationships is perceived to correspond closely to user's perception of the real world concepts. Moreover, with the inclusion of abstraction concepts such as aggregation, generalization/specialization and categorization in to the EER model, the semantic capability of the EER model is greatly enhanced. The presence of these abstractions also permits the schema designer to build up complex objects from primitives and offers a uniform approach to the expansion of the schema. As for the representational model, the relational model is selected because of its popularity and the -existence of a standard database language (i.e. SQL). VKQL is a complete query language comprising both the Visual Knowledge Definition Language (VKDL) and the Visual Knowledge Manipulation Language (VKML). A conversion algorithm is provided to translate VKQL queries to the Standard SQL equivalents. VKQL also does automatic integrity checking to ensure that the integrity constraints depicted on the EER schema is satisfied after each operation. As for the expressive power of VKQL, the language has been proven to be relationally complete. The main contributions of this research are the design and development of a relationally complete visual query language for user-database interaction and the proposal to use two separate models at the conceptual level, namely the knowledge model (i.e. EER model in this case) and the representational model (i.e. relational model in this case), to achieve knowledge abstraction. Other contributions include the design of an algorithm for converting VKQL queries to Standard SQL queries; and the design and development of a CASE tool to facilitate the specification of the EER schema.