## Multilingual Ontology Alignment: A New Frontier A new study presents a system for cross-lingual ontology alignment, based on the use of cosine similarity between vector embeddings. The goal is to identify and link equivalent concepts expressed in different languages, a crucial problem for data integration and the creation of global knowledge graphs. The proposed system stands out for the contextual enrichment of ontology entities, obtained through innovative description techniques. A fine-tuned multilingual transformer model generates high-quality embeddings. Cosine similarity is used to identify positive ontology entity pairs, with threshold-based filtering to retain only the most similar entities. The system was evaluated on the OAEI-2022 multifarm track dataset, achieving an F1 score of 71%, a 16% improvement over the best baseline system. This result suggests that the proposed alignment pipeline is able to capture the subtle cross-lingual similarities between entities. Knowledge graphs have become an essential tool for representing and managing knowledge in various domains. The ability to align multilingual knowledge graphs opens new perspectives for sharing and integrating information globally, overcoming language barriers.