Research Publications

Semantic Topology in Multilingual Knowledge Graphs: The Aevum Approach to Cross-Lingual Entity Resolution

This paper introduces a novel framework for resolving entity ambiguity across 140+ languages using graph neural networks augmented with cultural context vectors. Our method achieves a 94.7% accuracy in cross-lingual entity linking, outperforming existing baselines by 12.3%. We demonstrate applications in preserving endangered language knowledge and reducing bias in global information retrieval.