From Traditional to Innovation: Large Language Models in Fisheries Data Extraction

The fisheries industry is crucial for the global economy, providing essential food resources and livelihoods. However, accurate data collection from fishermen’s narrative reports is challenging due to their unstructured nature and low technological literacy among fishermen. This study explores the application of Large Language Models (LLMs), specifically the Merak-7B model, to automate data extraction from these reports. Due to resource limitations, Claude AI was used to generate narrative examples based on predefined prompts, reflecting potential reports by fishermen. Additionally, the study conducts a comparative analysis between Merak-7B and Claude AI to assess accuracy and reliability in data extraction. The Merak-7B model demonstrates high accuracy in converting unstructured text into structured data, thereby improving data collection efficiency and reliability. The results indicate that Merak-7B achieved an accuracy rate of 91% in extracting key data points from fishermen’s reports using 10 narrative examples as samples. This approach addresses the technological challenges faced by fishermen and supports better decision-making in fisheries management through enhanced data quality. The findings suggest that LLMs can significantly streamline data collection processes, aiding in sustainable fisheries management.