OntView: A Clearer Lens for Understanding Complex Ontologies

TLDR: OntView is a new open-source ontology viewer designed to simplify the visualization of complex knowledge structures. It uniquely displays inferred knowledge, including previously hidden elements like anonymous classes and General Concept Inclusions (GCIs). To combat information overload, OntView offers smart summarization techniques based on concept importance and provides interactive controls for navigating large ontologies, making them more accessible and understandable for users.

In the world of knowledge management and computer science, ontologies serve as structured frameworks for organizing vast amounts of information. They define concepts and their relationships, helping computer systems understand and process data more effectively. However, a significant challenge has always been the lack of effective tools to visualize these complex structures in a way that is easy for humans to understand without being overwhelmed.

Many existing ontology editors and viewers struggle to graphically represent ontology structures meaningfully, making it difficult for users to grasp the intricate dependencies and properties within large knowledge frameworks. This often limits the ability of knowledge engineers to identify modeling errors and developers to reuse existing ontologies.

Introducing OntView: What You See Is What You Meant

A new ontology viewer called OntView aims to address these challenges by providing an intuitive visual representation of ontology concepts and their formal definitions through a user-friendly interface. Built upon the use of a Description Logic (DL) reasoner, OntView follows a “What you see is what you meant” approach, displaying the actual inferred knowledge rather than just what was explicitly stated. This means users can see the complete, reasoned understanding of the ontology.

One of OntView’s standout features is its ability to visualize General Concept Inclusions (GCIs), which are complex logical statements that define relationships between concepts. This capability is largely absent in other visualization tools. To combat information overload, especially with large ontologies, OntView offers several ways to simplify the view:

Creating ontology summaries by assessing the importance of concepts using various algorithms.
Focusing the visualization on specific elements between two given classes.
Allowing users to dynamically hide or show different branches of the ontology without losing the underlying meaning.

OntView has been released with an open-source license, making it available to the entire community. For more in-depth technical details, you can refer to the full research paper: OntView: What you See is What you Meant.

Addressing Limitations of Existing Tools

The developers of OntView compared it against several prominent ontology visualization tools, including OWLViz, OntoGraf, WebVOWL, OWLGrED, and KC-Viz. While these tools offer various functionalities, they often fall short in key areas. For instance, many cannot visualize properties, instances, or anonymous classes, or they lack interactive features for navigating large ontologies. Some have steep learning curves or suffer from visual clutter as ontologies grow.

OntView distinguishes itself by offering:

Meaningful Semantic Visualization: It uses a rich visual language and a DL reasoner to show the true semantics, including anonymous classes (concepts without explicit names) and GCIs, which are often hidden in other tools.
Interactive Navigation: Users can explore the ontology by hiding and expanding elements dynamically, with continuous feedback on the visible and hidden parts. It also allows incremental exploration and searching for terms.
Information Overload Counter-measures: Beyond the summarization techniques mentioned earlier, OntView can automatically generate simplified views based on concept relevance, displaying only the most important elements.

How OntView Visualizes Ontologies

OntView organizes the graphical elements of an ontology into horizontal levels, representing the hierarchical distance of concepts from a universal “top” concept. More general classes appear on the left, and more specialized classes move towards the right. Key visual elements include:

Primitive Classes: Basic concepts shown as light gray rectangles.
Anonymous Classes: Concepts without explicit names, used in complex expressions.
Defined or Equivalent Classes: Concepts with specific conditions, highlighted with equivalent classes in green.
Class Disjointness: Indicated by a “D” marker and red lines, showing concepts that cannot share instances.
Object and Data Properties: Marked with a “P” symbol, revealing lists of properties when clicked.
IsA Connectors: Solid blue lines showing direct hierarchical “is-a” relationships.
Dashed Connectors: Dashed lines indicating indirect hierarchical relationships where intermediate nodes are hidden.
Range Connectors: Light blue lines linking properties to their defined ranges.
Property Hierarchy Connectors: Black lines showing sub-property relationships.

These visual cues, combined with tooltips and contextual menus, provide comprehensive information about each ontological element.

Advanced Visualization Techniques

OntView’s ability to handle GCIs and anonymous classes is a core innovation. Instead of trying to classify all anonymous classes (which can be computationally intensive for large ontologies), OntView intelligently classifies only those that will be displayed to the user. This involves building an initial graph of named classes and then precisely locating anonymous class expressions within that hierarchy using subsumption tests.

Users can also control the level of detail by selecting a specific fragment of the ontology to be detailed, ensuring they only see the relevant anonymous class expressions within that scope.

Smart Summarization and Expansion Control

Recognizing that showing an entire large ontology can be overwhelming, OntView incorporates smart summarization methods. It uses different algorithms to assess the importance of concepts and select the most relevant ones to display initially. These methods include:

Key Concept Extraction (KCEn): Measures importance based on cognitive, statistical, and topological factors.
PageRank/RDFRank: Graph centrality measures applied to the class taxonomy, also considering anonymous classes.
Custom Summary: Allows users to manually select specific concepts they are interested in visualizing.

For very large ontologies, OntView also provides fine-grained control over expansion and collapse. Users can define the percentage of nodes they want to show or hide in each step, with policies that prioritize relevance, more general descendants, or more specific descendants. A sliding bar also offers a global control for expanding the ontology based on overall relevance, providing an “eye-hawk overview” to identify emerging subdomains.

Under the Hood

OntView is developed in Java, utilizing the OWLAPI for handling ontologies and interfacing with Description Logic reasoners like Openllet and HermiT. JavaFX is used for the graphical interface, and Sugiyama’s algorithm helps minimize connector crossings for clearer diagrams. The project is open-source and available on GitHub.

Also Read:

Conclusion

OntView represents a significant step forward in ontology visualization. By offering an intuitive and powerful tool that displays inferred knowledge, handles complex expressions like GCIs and anonymous classes, and provides intelligent summarization and navigation controls, it makes large and complex ontologies more accessible and understandable for both creators and users. Future work includes systematic user evaluation, extending functionality to display individual instances, and developing a Protégé plugin to broaden its adoption.