Open Information Extraction with LLM for the Portuguese Language

Abstract

In this study, we investigate the application of Large Language Models (LLMs) for Open Information Extraction (OpenIE) in the Portuguese language. While most OpenIE methods have been developed with a focus on the English language, few works in the literature explore multilingual and cross-linguistic scenarios. Although there is a growing interest in OpenIE methods for Portuguese, the use of LLMs specifically focused on OpenIE in this language remains underexplored. We analyze the feasibility of incorporating both open and commercial LLMs using few-shot prompt engineering for OpenIE in Portuguese. We provide a detailed analysis of the performance of these LLMs in OpenIE tasks, demonstrating that they achieve performance metrics comparable to state-of-the-art systems. Additionally, we refine and release an open LLM for OpenIE, named PortOIE-Llama, which outperforms commercial LLMs in our experiments. Our results highlight the potential of LLMs in OpenIE tasks in Portuguese and suggest that further refinement and fine-tuning of larger models can further enhance these outcomes.