Zero-Shot End-to-End Relation Extraction in Chinese: A Comparative Study of Gemini, LLaMA and ChatGPT
Shaoshuai Du, Yiyi Tao, Yixian Shen, Hang Zhang, Yanxin Shen, Xinyu Qiu, Chuanqi Shi
TL;DR
This study tackles zero-shot end-to-end relation extraction in Chinese by comparing three large language models—ChatGPT, Gemini, and LLaMA—across accuracy and latency. It introduces a semantic matching-based evaluation to robustly compare outputs from different prompts and architectures on the DuIE 2.0 dataset. Results show ChatGPT (especially GPT-4-turbo) achieving the best overall accuracy, Gemini offering the fastest inference, and LLaMA lagging in both accuracy and speed, highlighting clear trade-offs between precision and efficiency. The findings inform practical deployment decisions for real-time versus high-precision Chinese RE tasks and point to directions for improving model adaptability and evaluation in Chinese NLP.
Abstract
This study investigates the performance of various large language models (LLMs) on zero-shot end-to-end relation extraction (RE) in Chinese, a task that integrates entity recognition and relation extraction without requiring annotated data. While LLMs show promise for RE, most prior work focuses on English or assumes pre-annotated entities, leaving their effectiveness in Chinese RE largely unexplored. To bridge this gap, we evaluate ChatGPT, Gemini, and LLaMA based on accuracy, efficiency, and adaptability. ChatGPT demonstrates the highest overall performance, balancing precision and recall, while Gemini achieves the fastest inference speed, making it suitable for real-time applications. LLaMA underperforms in both accuracy and latency, highlighting the need for further adaptation. Our findings provide insights into the strengths and limitations of LLMs for zero-shot Chinese RE, shedding light on trade-offs between accuracy and efficiency. This study serves as a foundation for future research aimed at improving LLM adaptability to complex linguistic tasks in Chinese NLP.