Researchers have developed a new open-weight pipeline for multilingual joint entity-relation extraction, designed to build signed, temporal knowledge graphs from large news corpora. This system combines named-entity recognition with a linking cascade to map mentions to Wikidata identifiers, followed by a mixture-of-experts model for extracting directed relationships. The pipeline demonstrated high accuracy in a spot-check against a gold standard and was validated through case studies in Austria and Poland, reconstructing political party lifecycles and uncovering complex economic and governance networks. AI
IMPACT This framework offers a robust, replicable method for cross-national computational social science, enabling deeper analysis of political and economic networks.
RANK_REASON The cluster contains a research paper detailing a new methodology and pipeline for entity-relation extraction.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →