RUIE: Retrieval-based Unified Information Extraction using Large Language Model
This work addresses the problem of computational inefficiency and poor generalization in UIE for AI researchers and practitioners, offering a trainable retrieval plugin that is incremental over existing LLM methods.
The paper tackles the challenge of generalizing unified information extraction (UIE) to unseen tasks with large language models (LLMs) by proposing RUIE, a retrieval-based framework that uses in-context learning and a novel demonstration selection mechanism, achieving average F1-score improvements of 19.22 and 3.22 over baselines on eight datasets.
Unified information extraction (UIE) aims to extract diverse structured information from unstructured text. While large language models (LLMs) have shown promise for UIE, they require significant computational resources and often struggle to generalize to unseen tasks. We propose RUIE (Retrieval-based Unified Information Extraction), a framework that leverages in-context learning for efficient task generalization. RUIE introduces a novel demonstration selection mechanism combining LLM preferences with a keyword-enhanced reward model, and employs a bi-encoder retriever trained through contrastive learning and knowledge distillation. As the first trainable retrieval framework for UIE, RUIE serves as a universal plugin for various LLMs. Experimental results on eight held-out datasets demonstrate RUIE's effectiveness, with average F1-score improvements of 19.22 and 3.22 compared to instruction-tuning methods and other retrievers, respectively.