Ski: A Semantics-Knowledgeable Interface

The Semantics-Knowledgeable Interface (Ski) is designed to provide a powerful but usable interface to databases. To do this, it combines the expressiveness of semantic models, the broad communications “bandwidth” of graphics, and a novel approach to the representation and manipulation of database schemas. 1. lIltroduction This paper describes a database interface which is under construction. Ski, the Semantics-Knowledgeable Interface, is built on top of an existing semantic database implementation called the Semantic Database Constructor (see [FKM84]). Ski is intended as a naive interfade, but allows access to a full semantic data deflnition and manipulation language called Semdal (see [KS34]). The implementation, language, and interface together form a semantic DBMS called Sembase. Sembase is based on a semantic model. Semantic modeling in general is discussed in [KB84b]; briefly, semantic database models use a semantic network approach in providing data modeling constructs which ere more expressive than traditional data models (i.e., relational, hierarchical, network). A semantic model rovides facilities for detiing non-atomic objects P abstraction) through attributes and for constructing type/subtype hierarchies (generalization). The model under1 ‘ng Sembase is described in more detail in [KMBZj?and [KMEMa]. The next section provides a brief overview of Ski. Section 3 describes the screen layout of Ski. Due to space limitations a simple example, rather than a complete system description, is given in Section 4. This work was supported by IBM under a Faculty Development Award (1984). Permission to copy without fee 011 or pan of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the VLDB copyright notice and the title of the publicaiion and its date appear, and notice tk given that copying is by permission of the Very Large Data Base Endowment. To copy otherwise, or to republish, requires a fee and/or special permission from the Endowment. Proceedings of the Tenth Intemationel Conference on Very Large Data Ssses. 2. Ovemiew A few researchers have examined the issue of using a semantic database as the framework for a graphicsbased user interface (e.g., MPS [Wi33]; and GUIDE [WK62]). A (non-graphics) interface based on a semantic model is described in [Mo83]. In [SK821 a graphics-based relational browser is describe& while it does not use a semantic model, it does support powerful window and icon manipulations. Semantic models are both expressive and complex. The challenge in creating a semantic interface lies in making semantic modeling a manageable tool. Ski does not rely on a statically defined graphical layout of the schema. Instead, we attempt to use the semantic relationships inherent in the schema definition to dynamically drive the graphical representation. The portions of the database to be displayed and the types of relationships depicted vary with each user’s area of interest, as indicated through their use of the SKI operators. Ski uses a formated screen to simplify the representation of semantic schemas and supports semanticallymotivated operators. While interacting with Ski, the user uses these operators to create a session view. The user selects schema components of interest and peruses the schema for related information. The unique feature of Ski is that this perusing is not performed navigationally, but semantically. The user may also reconfigure or access the database using the Ski operators. Ski provides three classes of operators for perusing the schema while creating a session view. Ski allows the user to: (1) explore the direct semantic relationships in the schema by examining attributes, subtypes, constraints, end parent types, (2) determine how two schema components are related by locating and displaythe paths through the schema which connect them, 8 . 3 view the update entimm.ent of a specific schema component with respect to a particular update operation by displaying the schema components which are likely to be affected by performing that operation on the speciAed component. These last two sorts of perusing are actually quite complex, due to the existence of subtypes. The predicates which deAne the membership of subtypes may involve schema components which structurally are from very diierent parts of the schema. As an example, a smart student may be defined as one who gets an A in a course such that the professor who teaches that course is known to give tough grades. The information about a professor’s grading history may not be “near” the student’s information schematically, but the two are very closely rel,ated. Ski consists of six window subsystems. First, the user enters the Hello subsystem, which provides a brief Singapore, August, 1994