Enhancing Relation Extraction Through Augmented Data: Large Language Models Unleashed

Relation extraction models trained on scarce datasets often exhibit poor performance. The majority of datasets for relation extraction suffer from scarcity, resulting in decreased overall model performance-especially for infrequently encountered relations during training. To alleviate this problem, we present a methodology for augmenting data using large language models by applying in-context learning for relation extraction tasks. Our results reveal that data augmented using different language models can yield distinct effects on relation extraction tasks. Additionally, we compared the performance of the augmented data with other state-of-the-art approaches for data augmentation and conducted a comprehensive analysis of the results. Our findings demonstrate that large language models can produce significantly improved augmented data without the need for fine-tuning, to be utilized in enhancing relation extraction models.