Achieving a robust and universal semantic representation for action description remains a key challenge check here in natural language understanding. Current approaches often struggle to capture the complexity of human actions, leading to imprecise representations. To address this challenge, we propose a novel framework that leverages multimodal le