Enabling Traceable Eparticipation in Legislative Processes Using Akoma Ntoso
Year: 2026
Type: conference-paper
Abstract: This paper presents a public consultation platform for legislative bills encoded in Akoma Ntoso (AKN) XML that keeps every comment traceable to the cited legal text. Inspired by the crowdlaw paradigm-the use of technology to involve citizens in lawmaking-the platform lets citizens anchor feedback to specific passages, propose modifications with a separate rationale field, and search across documents using semantic similarity. The analytics pipeline builds interpretable signals from deterministic aspect extraction, Natural Language Inference (NLI)-based stance inference assembled into Perspectivized Stance Vectors (PSV), thematic clustering via$k$-means, and Principal Component Analysis (PCA) on comment-anchor difference embeddings. Aggregates retain explicit back-links to comment identifiers and target segments, enabling transparent review without decoupling feedback from the legislative structure. We describe the architecture and data model, detail the NLP pipeline, and report preliminary evaluation on three Italian parliamentary bills (146 comments). Results show that SBERT anchoring substantially outperforms random assignment but is surpassed by a TF-IDF lexical baseline, reflecting high terminological overlap between comments and legal provisions. Domain-general NLI exhibits a strong neutral bias on Italian legal text (24-49% accuracy); a hybrid NLI+LLM pipeline raises accuracy to 74-94% (macro-F1 0.63-0.71).
Authors: Simone Vagnoni, Monica Palmirani
Institution: University of Bologna
Language: English
Subfield: Law
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7214734652
