UC Santa Cruz

The transformation of data from one schema, called the source schema, to another schema, called the target schema, is the focus of data exchange. The primary challenge in data exchange is to develop methods that transfer data from a source schema to a target schema using schema mappings that specify the relationship between the two schemas so that the resulting data accurately reflects the source data. Over the past two decades, extensive research has been conducted on data exchange, starting from the original formalization of the problem for relational schemas. While temporal databases have been extensively studied for many years, the exploration of temporal data exchange is a relatively recent development. We investigate the notion of temporal schema mappings specifying the relationship between two temporal schemas. Temporal schema mappings contain at least one temporal relation symbol and use Allen’s relations to describe the relationship between temporal variables. Temporal data in such databases can be represented either through time intervals (concrete time) or time points (abstract time). Taking this into account, we design a chase algorithm for temporal data exchange settings with multipletemporal variables in the context of concrete time. We demonstrate that this algorithm produces universal solutions, provided that it does not fail. Furthermore, we investigate the relationship between universal solutions in the context of concrete time and universal solutions in the context of abstract time. We find that challenges arise even when temporal schema mappings involve a single temporal variable, but we also identify scenarios where these challenges can be overcome.

Different applications may use data structured in various formats, such as relational data or RDF. Depending on the applications, the target data may need to adhere to a relational schema, while others require it to conform to an RDF-expressed domain ontology. Our research also focuses on the following problem: Given a set of temporal schema mappings, how to exchange data with temporal information from a relational source schema into a target RDF-expressed ontology, so that we can enrich both the data and the ontology with temporal information from the relational sources? To address this challenge, we design a domain-independent algorithm that materializes target RDF data via a version of data exchange. This algorithm ensures the enrichment of both the data and the ontology with temporal data obtained from the sources.

The aforementioned work assumes that the temporal schema mappings have already been derived. Earlier studies on relational data exchange addressed the problem of active learning of (non-temporal) global-as-view (GAV) schema mappings. In the last part of this dissertation, we initiate an investigation on the automatic derivation of relational-to-RDF temporal schema mappings (based on our earlier work on relational-to-RDF temporal data exchange). To achieve this, we design an active learning algorithm and validate it using the metadata generator iBench, which was originally developed to generate relational-to-relational schema mappings and their corresponding data examples. We enhance iBench with additional features, referred to as temporal iBench, to enable the generation of relational-to-RDF temporal schema mappings and their data examples. Additionally, we carry out a comprehensive experimental evaluation which demonstrates the effectiveness of our active learning algorithm.