Anmerkungen:
Tag der Verteidigung: 30.09.2023
Zusammenfassungen in deutscher und englischer Sprache
Beschreibung:
As the amount of data represented in knowledge graphs continues to grow, there is a strong demand for data exploration techniques that are accessible to non-expert users. Keyword search is a well-known and user-friendly paradigm that can be used as an alternative to structured query languages such as SPARQL. Based on a provided implementation that generates SPARQL queries for a given set of keywords and a target type, we design and implement a full-stack search application for knowledge graphs. To achieve comprehensive result sets we combine the results of multiple SPARQL queries and provide them with a ranking. In the interface, we also present information about the subgraphs represented by the SPARQL queries, providing semantic context for the results and increasing the transparency and trustworthiness of the search outcome. In a user evaluation of the search application running against the Mondial knowledge graph, we observed that participants were able to use the provided information to contextualize the results and, if necessary, adapt the search to get the information they wanted. We also uncovered further potential for improvement in both the search application and the query generation.