Schema extraction and visualization of linked open data | |
| Author | Dangol, Reshma |
| Call Number | AIT Thesis no.IM-18-03 |
| Subject(s) | Ontology Linked data Semantic Web Database managemnet |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Information Management, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. IM-18-03 |
| Abstract | Linked Open Data (LOD) is one of the core concepts of Semantic Web. Publishing data as LOD results in unexpected re-use of the data which in turn increases its value. However, there are still a lot of challenges remaining for LOD users. Identifying appropriate datasets that fits a specific requirement is still a strenuous task and requires significant amount of human effort. The flexibility offered by Linked Data in terms of mixing up several ontologies/vocabularies in itself creates a challenge as it leads to difficulty in understanding the schema of the dataset. Users need to rely on SPARQL queries to explore and understand new datasets, which means that people without SPARQL knowledge are unable to consume LOD. Additionally, most of the existing ontology visualization tools are suitable only for ontology experts and only few of the tools can represent all key elements of the ontology. In this study, we have tried to overcome the aforementioned challenges by developing a system that is capable of extracting summary schema information from Linked Data dumps and visualizing the extracted information in a highly interactive interface. LD-VOWL is an appreciable effort with similar goals. However, the extracted details are quite limited and the graph visualization becomes completely unusable for dense schema graphs. Our tool VizLOD has been developed as an improvement on the works of LD-VOWL. VizLOD is capable of extracting various inferable ontological details and instance level information. The issues of scalability often found while dealing with node-link graphs has also been well addressed. The graph simplification techniques and interactivity in VizLOD makes exploring even the most dense schema graph an easy task. The evaluation of the accuracy of schema extraction was not as straightforward as initially expected. We realized that even though we may have ontology file for the LOD dataset to compare the extracted result with, we cannot be sure that the ontology is 100% correct for the dataset. Often the data in a LOD dataset do not conform to all the ontological details specified in its schema file. Furthermore, the extracted schema information that is not present in the schema file may not necessarily be incorrect. The results of comparative user evaluation between LD-VOWL and VizLOD have confirmed our hypothesis that the proposed interactivities in VizLOD does improve user experience and enables easier understandability of Linked Datasets. Even the novice Linked Data users were able to easily use the tool and understand new datasets in more detail. The experiments have also proved that VizLOD is a scalable tool capable of handling different levels of complexity. |
| Year | 2018 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. IM-18-03 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information Management (IM) |
| Chairperson(s) | Vatcharaporn Esichaikul; |
| Examination Committee(s) | Guha, Sumanta;Janecek, Paul; |
| Scholarship Donor(s) | Asian Institute of Technology Fellowship; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2018 |