Introducing knowledge footprints: Enhancing knowledge graph data retrieval with visualisations for improved geoscience collaboration
Summary
The EO4GEO Body of Knowledge (BoK) serves as a hierarchical structure of describing concepts in the domain of earth observation and geoinformation. Due to this nature the EO4GEO BoK provides an interesting data source for tagging and visualising organisational and personal knowledge. The hypothesis suggests that by publicly disclosing and simplifying access to information about an entity’s specific expertise or knowledge, it becomes easier to identify experts, thereby improving collaboration within a particular field.
This study presents how the already existing relational nature of describing BoK concepts is transformed to a RDF graph dataset following an ontology for describing bodies of knowledge. In addition, this graph dataset is then semi-automatically enriched with expertise annotations, which are created by extracting EO4GEO concepts from research papers by natural language processing tools. These expertise annotations and the hierarchical and relational structure of the graph dataset creates the EO4GEO knowledge graph. With this knowledge graph, it becomes possible to ask questions that suggest which individual or organisation has specific expertise on a topic. These topics match the concepts in the EO4GEO body of knowledge.
This study further shows how visualisations can be leveraged and how they provide extra context in knowledge graph data retrieval through question and answering through visualisations. These visualisations are called knowledge footprints. They are created for the purpose of representing, promoting and retrieving someone’s expertise. This study introduces a user-evaluated website that combines the EO4GEO knowledge graph, SPARQL, JavaScript and the D3.js library to interactively create these knowledge footprints. In addition, this website introduces potential applications for knowledge footprints.
The evaluation suggests that knowledge footprints do offer an interesting approach to visualise knowledge in our geospatial domain. While the evaluators see value in using knowledge footprints to identify collaborators with specific expertise, knowledge footprints likely won’t entirely replace traditional search methods. Collaboration often involves certain human factors like reputation and familiarity that go beyond domain knowledge.