From Parliamentary Rhetoric to Enacted Law: An NLP Pipeline for Semantic Auditing of the Greek Legislative Process

Abstract

The Greek legislative framework is characterized by intricate cross-referencing, frequent amendments, and limited machine-readable access, hindering transparency and civic engagement. Traditional bulk-archiving approaches are computationally expensive and fail to capture political relevance. We present a multimodal computational pipeline that bridges parliamentary discourse with enacted legislation. Applying Natural Language Processing (NLP) to 2025 Hellenic Parliament transcripts, we extracted 534 unique law citations and used debate frequency as an empirical signal to identify politically salient laws. A headless browser architecture enables automated acquisition of official Government Gazette documents despite anti-scraping barriers. Using Large Language Models (LLMs), we conduct a semantic audit of legislative quality. Our analysis reveals an "Illusion of Simplicity", where laws framed as simplifications exhibit high structural complexity and ambiguity. A typology of 312 ambiguity instances shows that 45 percent stem from vague terminology and 25 percent from deferred executive delegation. We introduce the Political Discrepancy Index (PDI), evaluating alignment between ministerial promises and enacted law. Across three high-frequency laws (4808/2021, 4412/2016, 4662/2020), the dominant outcome is Deferral, with commitments shifted to future Ministerial Decisions. Cross-reference network analysis confirms a highly entangled legal system, with foundational provisions among the most frequently amended. The pipeline produces a semantically linked dataset and an interactive auditing interface for scalable analysis of legislative processes.