SWRL rules are increasingly being used for knowledge representation in OWL ontologies. As the size of these rule bases increases, users face the problem of understanding and managing the increasing amounts of knowlege represented by the rules. To support the management of the resulting complexity, rule management techniques are needed. These techniques include the ability to categorize rules, to provide simplified explaination of rules, and to provide high-level abstractions of rule base structure. In particular, rule paraphrasing and rule visualization can help nonspecialists to understand the meaning of complex rules. Abstraction of common patterns in rule bases can also enable automatic categorization of rules into related groups. These categories can also be used as a basis of rule elicitation tools that guide users when entering new rules.
To tackle this problem, we have developed rule management and elicitation software called Axiomé [3]. Axiomé was developed as a plug-in for the popular Protégé-OWL ontology development environment and provides an array of tools for managing SWRL rule bases in OWL ontologies. These tools supports visual rule base exploration, automated rule categorization, rule paraphrasing, and rule elicitation functionality with the goal of facilitating the management of large SWRL rule bases.
Axiomé was developed as a Protégé-OWL plug-in and has five main functional areas. These functionalities are available as sub-tabs within the plug-in: (1) a Rule Graph tab that provides a graph structure to browse and explore SWRL rule base and rule relationships; (2) a Rule Visualization tab to visualize individual rules; (3) a Rule Paraphrasing tab that displays an English-like text explanation for each rule; (4) a Rule Categorization tab to automatically categorize rules in a rule base; and (5) a Rule Elicitation tab that provides a graphical templates to acquire new rules based on analysis of existing rules in a rule base. A Rule Browser component is permanently displayed to show a tree-table representation of the SWRL rules in an ontology. This tree-table enables users to explore the rule base and launch any of five sub-tabs for the rule or group being explored.
The Rule Graph tab provides a graphical representation of a SWRL rule base (Figure 1). Each rule in the rule base is presented as a node, and directed edges between two rules indicate the SWRL atoms shared by the rules and the dependency direction. A number of graphical layouts are supported, using the JUNG visualization framework [2] to produce optimal graph layouts. Search functionality is also provided and rules partially matching the search term can be highlighted in the graph. Rule groups and the types of dependencies between rules which are based on the common SWRL atoms between them, can also be indicated visually. Cyclical dependencies between rules can be found and visually highlighted in the graph.
The Rule Visualization tab (Figure 2) allows individual rules to be visualized as a tree structure. Trees are generated by depth-first search for each variable chain in the rule, with more prominent atoms being placed near the root of each tree. A number of heuristics are employed to ensure that the most important clauses in a rule are given more prominence [1].
The Rule Paraphrasing tab (Figure 3) uses a similar approach to build a tree structure for each rule and then uses additional heuristics to generate understandable English paraphrases of that rule [1].
The Rule Categorization tab uses the data structure that is generated for the rule visualization and paraphrasing tabs to automatically group the rules with a similar syntactic structure [1]. It then graphically displays the results of this grouping. These groupings can then be used in the Rule Graph tab when exploring the rule base.
The Rule Elicitation tab (Figure 4) provides graphical rule templates to facilitate acquisition of rules. It generates these templates using the syntactic structures generated for rule groupings. Users can select an appropriate rule group and then generate a graphical acquisition template to enter rules with the structure of other rules in that group.