P4E: Few-Shot Event Detection as Prompt-Guided Identification and Localization

We propose P4E, an identify-and-localize event detection framework that integrates the best of few-shot prompting and structured prediction. Our framework decomposes event detection into an unstructured identification task and a structured localization task. For the unstructured identification task, we leverage prompting to elicit knowledge from pretrained language models, allowing our model to adapt to new event types quickly. We then employ a type-agnostic sequence labeling model to localize the event trigger conditioned on the identification output. This heterogeneous model design allows P4E to make fast adaptation without sacrificing the ability to make structured predictions. Our experiments demonstrate the effectiveness of our proposed design, and P4E achieves the new state-of-the-art on few-shot entity detection across multiple datasets.