Visual Justifications for Ontologies
Lead Research Organisation:
University of Brighton
Department Name: Sch of Computing, Engineering & Maths
Abstract
Ontologies are a way of reasoning about data in an efficient manner. Ontologies are increasingly prevalent in a range of applications, including the Semantic Web, medicine and law. The development and maintenance of ontologies are skilled tasks requiring knowledge of logical reasoning and symbolic notations. However, the wide variety of stakeholders for each ontology may not have the necessary skill set to perform ontology engineering effectively. Given the critical systems in which ontologies are used, it is of paramount importance that the ontologies encode exactly the information intended.
Ontologies containing errors, called incoherent ontologies, undergo debugging or repair by an ontology engineer. Extant ontology reasoners provide a justification for the incoherence. However, interpreting the justification is a non-trivial and difficult task. Even if the engineer understands the domain of the ontology, for example medicine, and the symbolic notation in which the justification is represented, they could still struggle to debug the ontology. It is especially difficult to debug the ontology without unintentionally removing intended behaviour.
This project will use concept diagrams to visualise justifications to reduce the burden on the ontology engineer. Using concept diagrams will help both the understanding of the problem and suggest appropriate repairs to the ontology. The project will provide a number of different visualisations of common bugs in ontologies and empirically test the effectiveness of each. Through this process, the project team will be able to develop effective visual justifications. Using real-world examples of ontologies for data privacy supplied by the project partner HERE (a Nokia company) the visual justifications will then be tested against equivalent symbolic and natural language justifications.
Ontologies containing errors, called incoherent ontologies, undergo debugging or repair by an ontology engineer. Extant ontology reasoners provide a justification for the incoherence. However, interpreting the justification is a non-trivial and difficult task. Even if the engineer understands the domain of the ontology, for example medicine, and the symbolic notation in which the justification is represented, they could still struggle to debug the ontology. It is especially difficult to debug the ontology without unintentionally removing intended behaviour.
This project will use concept diagrams to visualise justifications to reduce the burden on the ontology engineer. Using concept diagrams will help both the understanding of the problem and suggest appropriate repairs to the ontology. The project will provide a number of different visualisations of common bugs in ontologies and empirically test the effectiveness of each. Through this process, the project team will be able to develop effective visual justifications. Using real-world examples of ontologies for data privacy supplied by the project partner HERE (a Nokia company) the visual justifications will then be tested against equivalent symbolic and natural language justifications.
Planned Impact
The beneficiaries of this work are primarily ontology engineers. The ability to debug incoherent ontologies using visual justifications will alleviate some of the burden of the difficult task of ontology repair. By making the process of ontology design and maintenance more accessible and precise, engineers can have more confidence in their ontologies and thus all stakeholders will benefit. The main impacts will therefore be in industrial applications including, but not limited to, our project partner HERE. Avoiding the misuse of data is a legal requirement for HERE and they use ontologies to ensure data is managed properly. Guaranteeing the consistency and correctness of the ontology, with respect to the situation modelled, is thus very important for HERE. Other companies with similar requirements can benefit from the visual justifications, enabling more confidence in ontologies throughout industry.
Visualisation helps the various stakeholders using an ontology communicate more effectively. Since reasoning using symbolic notations requires significant mathematical training, non-mathematically trained, but domain expert, stakeholders such as lawyers may struggle to communicate precisely with ontology engineers who may not be domain experts. By presenting information in a visual, yet precise and unambiguous manner, non-mathematical stakeholders can adequately convey the requirements and subtleties of the domain. Within the context of justifications, where more than one repair is possible, visualisation will enable the domain expert to suggest which repairs will be preferable. For example, some repairs may remove desirable behaviour from an ontology.
Visualisation helps the various stakeholders using an ontology communicate more effectively. Since reasoning using symbolic notations requires significant mathematical training, non-mathematically trained, but domain expert, stakeholders such as lawyers may struggle to communicate precisely with ontology engineers who may not be domain experts. By presenting information in a visual, yet precise and unambiguous manner, non-mathematical stakeholders can adequately convey the requirements and subtleties of the domain. Within the context of justifications, where more than one repair is possible, visualisation will enable the domain expert to suggest which repairs will be preferable. For example, some repairs may remove desirable behaviour from an ontology.
People |
ORCID iD |
| Peter Chapman (Principal Investigator) |
Publications
Hou T
(2016)
Measuring perceived clutter in concept diagrams
Molebatsi K
(2022)
Implementation Planning for Integrating Depression Screening in Diabetes Mellitus and HIV Clinics in Botswana.
in Global implementation research and applications
Rodgers P
(2015)
Visualizing Sets with Linear Diagrams
in ACM Transactions on Computer-Human Interaction
Stapleton G
(2018)
Graph Structures for Knowledge Representation and Reasoning
T. Hou
(2016)
Antipattern Comprehension: An Empirical Evaluation
| Description | The research focussed on whether using concept diagrams to represent inconsistent information would be useful to both novice and expert users in the field of ontology engineering. It was found that: - using concept diagrams improved the understanding of novice users. Both the time and accuracy of identifying inconsistencies improved when using diagrams over using text. Further, merging information into a single diagram, rather than representing it using several diagrams, was more beneficial. These findings are based on a quantitative study involving over 60 participants. - for expert users, the findings were more nuanced. Based on a qualitative study involving 10 participants, both benefits and drawbacks were found in using concept diagrams to investigate inconsistent information. The benefits were similar to those of novice users: an increased level of confidence for some types of information. However, there were concerns over scalability: whether or not this approach would be appropriate for larger knowledge bases. |
| Exploitation Route | The findings may be taken forward by a careful implementation of concept diagrams into an ontology editing environment, once any changes to the diagrams' appearance have been made. We now know when concept diagrams are useful, and when they may be less useful, for identifying and repairing inconsistent information. A first focus would be on an implementation in the areas where concept diagrams have been shown to be effective, to produce a useful tool for ontology engineers. The types on information where concept diagrams are not, as yet, as effective, will require further empirical research to address. |
| Sectors | Digital/Communication/Information Technologies (including Software) |