A Scalable Workflow for Extracting Species Interactions From Ecological Literature Using Large Language Models

The contemporary ecological crisis calls for integration and synthesis of vast amounts of mostly scattered and often hard-to-extract ecological information that is dispersed across hundreds of thousands of scientific papers. Recent advancements in natural language processing (NLP) and the emergence of large language models (LLMs) offer a novel and promising approach to this persistent challenge. Here, we present a replicable and adaptable workflow that automates the extraction of a broad range of biological interactions among species from scientific literature using LLMs. We provide a demonstration of this workflow by performing a large-scale data extraction directly from a corpus of 83,910 scientific articles using OpenAI GPT-4o. Our approach successfully extracted a network of 144,402 interactions between 36,471 taxa. Performance analysis shows that the model exhibits a precision of 89.5% with a recall of 70.0%, demonstrating that a general-purpose LLM can perform complex ecological information extraction at scale for analysis and synthesis. Results depend on quality filtering choices and inherit biases present in the underlying literature, emphasizing the need for validation and quality control. Standardized outputs can be browsed, queried, or joined with existing interaction repositories and taxonomic backbones, facilitating downstream reuse and interoperability. Our case study shows that LLMs are capable of carrying out complex extraction tasks on key ecological data on a very large scale, paving the way for a multitude of potential applications in ecology and beyond.