Inductive Framework for Learning Concept Embeddings via Colexification Patterns

This research presents an inductive framework to learn scalable concept embeddings using cross-linguistic colexification patterns and textual attributes. Unlike previous transductive methods, this approach allows for the generation of embeddings for nodes not present in the initial training data.