typhon-rml: Modularized Declarative Knowledge Graph Construction for Flexible Integrations and Performance Optimization

Marco Grassi, Mario Scrocca, Alessio Carenini, Irene Celino

Keywords: Declarative Mappings, Knowledge Graph Construction, Mapping Languages

TL;DR: A method and a tool to modularize the declarative construction of knowledge graphs, allowing for a decoupled processing of input/output data sources and mapping rules

Abstract: Adopting declarative approaches for constructing knowledge graphs enhances the maintainability and reusability of schemas and data transformations from diverse data sources. However, a fully declarative description requires the user to encode specific details of the data integration process within the mapping rules, including how to extract the input data from specific data sources and how to load the result into the target ones. This aspect significantly burdens the developers of mapping processors, who must adhere to the mapping language features to transform heterogeneous data formats to RDF, while also facilitating efficient access to various input and output data sources. Additionally, considering the user's point of view, a tightly coupled approach for the declaration and execution of the entire construction process can affect the flexibility of reusing mapping rules for different data integration scenarios. In this paper, we address these challenges and propose a method to modularise the declarative construction of knowledge graphs, allowing for a decoupled processing of input/output data sources and mapping rules. We introduce the typhon-rml library to demonstrate this approach. Focusing on RML mapping rules as input, we showcase how the tool facilitates their reuse and customisation for various integration requirements. A preliminary qualitative evaluation is conducted in the context of a relevant scenario for smart traffic management.

Submission Number: 10