CL4DEE: A continual learning framework for document-level event extraction with mixture-of-experts

Key Points

• The framework enhances document-level event extraction by integrating continual learning techniques, fostering adaptability.
• Key findings show a notable increase in extraction accuracy, particularly in varied document types, by 15% compared to previous methods.
• The approach employs a mixture-of-experts model to leverage specialized nodes, optimizing performance on specific tasks in real time.
• This work highlights the potential for mixed learning paradigms to improve efficiency and accuracy in machine learning applications.