DIAVGEIA (meaning clarity in Greek) is the Greek Open Government national portal where most of the public sector administrative acts and decisions are published forming a huge collection currently containing more than 60 million documents. The information published at DIAVGEIA include document-based metadata (e.g., the signer/publisher of the act/decision) and unstructured documents (pdfs). The majority of these documents contain information about transactions/events (e.g., payments, hirings, appointments, project assignments) that occur in the frame of Public Administration and concern specific agents (citizens, businesses, public servants etc.). However, the actual transaction data is not adequately modeled at the metadata or provided only in an unstructured way through the pdf documents. This paper proposes an approach to extract and model the transactional data available in DIAVGEIA as Event Centric Knowledge Graphs (ECKG). The approach uses ECKG models to represent the data and proposes a process, that leverages AI/ML/LLM techniques in order to generate the ECKG from the documents. This approach considers events as first-class entities for knowledge representation and works towards adopting a data-centric event-native mindset in data management of the public administration compared to the existing document-centric paradigm. This in turn can be beneficial for the public administrations in order facilitate real-time decisions making but can also increase transparency and accountability.