Knowledge graphs have gained momentum in the past few decades, becoming evermore a key asset for data interoperability, management, analysis and exploitation. An aspect that plays an essential role in their uptake and use is the ease to construct them. There are several ways in which knowledge graphs can be constructed, ranging from ad-hoc scripting to declaratively defining the transformation rules.
The use of these declarative approaches enable a reusable, maintainable and understandable manner for a seamless knowledge graph construction. They rely on languages for expressing the transformation rules that, while widely used and adopted, they still lack some expressiveness for the increasing complexity of available data. Therefore, this thesis analyses the expressiveness of these languages, extracts and defines the requirements for constructing knowledge graphs with the current needs. In addition, it extends a well-known language with features to generate knowledge graphs enriched with annotations, following the latest developments in the area.
In order to facilitate the creation of the transformation rules for users with different backgrounds and expertise, this thesis proposes a spreadsheet-based approach to write them, providing a familiar environment and suppressing the need of learning the language's syntax. It also updates a user-friendly serialization with the latest additions from its target language, for users with more technical profiles. Both this approaches are supported by implementations able to interoperate with different languages.
Finally, this thesis evaluates the role that these declarative approaches can play in different tasks involved in the knowledge graph life cycle. More specifically, it assesses how they can be beneficial when refactoring the schema of knowledge graphs.
Overall, this thesis contributes to the understanding of the capabilities that declarative languages for knowledge graph construction and refactoring, while providing extended support for their creation and interoperability.