The European General Data Protection Regulation (GDPR) brings new challenges for companies, who must provide transparency with respect to personal data processing and sharing within and between organisations. Additionally companies need to demonstrate that their systems and business processes comply with usage constraints specified by data subjects. This paper first presents the Linked Data ontologies and vocabularies developed within the SPECIAL EU H2020 project, which can be used to represent data usage policies and data processing and sharing events, including the consent provided by the data subject and subsequent changes to or revocation of said consent. Following on from this, we propose a concrete transparency and compliance architecture, referred to as SPECIAL-K, that can automatically verify that data processing and sharing complies with the relevant usage control policies. Our evaluation, based on a new transparency and compliance benchmark, shows the efficiency and scalability of the system with increasing number of events and users, covering a wide range of real-world streaming and batch processing scenarios.