AbstractThe main aim of this thesis is to determine, via empirical study, whether diagrams can more effectively support user understanding than textual and symbolic logics. A further aim is to identify, via empirical study, how to choose between syntactically different diagrams when formulating logical axioms; this is a prerequisite to evaluating them compared to symbolic and textual notations.
Effectiveness is judged by measuring task performance.
To ensure practical applicability, our empirical studies focus on axioms arising in ontology engineering, which is a major application area for logics. People need to understand the axioms of which ontologies comprise and derive inferences from them. Concept diagrams provide an ideal diagrammatic logic for our studies, as they were designed specifically for ontology engineering. Ontologies can also be created using textual notations, such as the Manchester OWL syntax (MOS), and symbolic notations, such as Description Logic (DL). Concept diagrams will be empirically compared to MOS and DL. All our empirical studies focus on commonly required semantic properties that require axiomatization in ontologies.
The first contribution, addressing the second aim, provides insight into choices between syntactically different but semantically equivalent concept diagrams. We systematically identified three approaches to axiomatizing semantic properties using diagrams. We evaluated these approaches and concluded that avoiding explicit quantification, and representing the information purely diagrammatically, best supports task performance. As a result, users are guided towards avoiding more syntax and high visual complexity where possible.
The second contribution, addressing the main aim, suggests that concept diagrams are more effective than both MOS and DL for a range of tasks. Firstly, our empirical results suggest that, for six common axiom types, concept diagrams, allow people to perform simple tasks which involve understanding single axioms, more effectively than MOS and DL; for these tasks, we also found that MOS outperforms DL in some cases. Secondly, the empirical results are extended to include a broader range of more complex tasks, including user understanding of sets of axioms and their entailments. The results again suggest that concept diagrams are more effective than MOS and DL. Surprisingly, we also found that MOS did not outperformed DL for these more complex tasks.
Overall, the results of this research provide evidence that concept diagrams help people understand a range of axiom types and derive inferences from them. Therefore, they could play an important role in supporting logical reasoning in ontology engineering and more broadly. Also, the results provide evidence that diagrams allow people to make a large number of perceptual inferences and lead to improved task performance.
|Date of Award
|John Howse (Supervisor)