Back to portfolio

TelarKG

Overview

TelarKG is a Knowledge Graph capturing data from Chile's 2021–2022 Constitutional Process. It integrates heterogeneous sources — convention members, voting records, sessions, commissions, proposed articles and biographies — into a unified semantic model that supports rich, cross-cutting queries that would be impractical against the original relational silos.

The project was developed inside the Millennium Institute Foundational Research on Data (IMFD) in collaboration with researchers from Universidad de Chile and Universidad Católica. The public endpoint at telarkg.imfd.cl exposes both a query interface and a curated set of analytical views.

Technical architecture

• Data ingestion. Heterogeneous inputs (CSV exports, public APIs, institutional dumps and scraped HTML) are normalized in a Python ETL layer using pandas and small purpose-built parsers, then materialised as RDF triples following a custom ontology aligned with reusable vocabularies.
• Graph store. The dataset is served from MillenniumDB, a multi-model graph engine developed at IMFD that supports both RDF/SPARQL-style and property-graph workloads, optimised for path queries and labelled property graphs.
• Query layer. Application queries are expressed in SPARQL and in MillenniumDB's path query language; the most expensive ones are precomputed and cached so the dashboards stay responsive.
• Web front-end. A Django application provides the endpoint, authentication, query templating and the visual layer (network views, timelines, member profiles and voting heatmaps).
• Deployment. Containerised with Docker behind an Nginx reverse proxy; static assets are served separately to keep the Django process lean.

Research contributions

• A reusable ontology for legislative-process data (sessions, commissions, votes, proposed norms and their relations to convention members) that can be re-applied to similar constitutional or parliamentary processes.
• Empirical evaluation of MillenniumDB on real-world political data, providing feedback on graph-engine ergonomics for non-trivial analytical workloads.
• A public, queryable dataset that lowers the barrier for journalists, social scientists and citizens to perform their own analyses on the constitutional process.

Related publications

1. Angles, R.; Bro, N.; Donoso-Guzmán, I.; Luna, J. P.; Hogan, A.; Reutter, J.; Rosales-Méndez, H.; Toro, S. Telar and TelarKG: Data-Driven Insights into Chile's Constitutional Process. Communications of the ACM, 2024. doi
2. Angles, R.; Calisto, V.; Díaz, J.; Ferrada, S.; Hogan, A.; Pinto, A.; Reutter, J.; Rojas, C.; Rosales-Méndez, H.; Sarmiento, H.; Toussaint, E.; Vrgoč, D. TelarKG: a Knowledge Graph of Chile's Constitutional Process. GRADES-NDA Workshop, 2024. pdf
3. Vrgoč, D.; Rojas, C.; Angles, R.; Arenas, M.; et al. MillenniumDB: A Multi-modal, Multi-model Graph Database Engine. SIGMOD'24 (Demo Track).

Open TelarKG