Description
Scientific publications have enormous amounts of information and serve as the main pillar for advancing knowledge across various disciplines. Recently, many sectors and disciplines have been employing Deep Learning (DL) models due to their popularity. However, manually extracting DL method information from publications is becoming tedious with the ever-growing published literature. On the other hand, validating and verifying this information is a pivotal step for checking the reproducibility of the DL pipeline in scientific publications. In this work, we leverage the multimodal information (text, figures, tables, graphs, etc.) to automatically retrieve the method information of DL pipelines in scientific publications using the suite of open-source models, including Large Language Models (LLMs) and computer vision models. We will present the initial results from the text modality of DL method information from biodiversity scientific publications drawn using open-source LLMs.
| Type of Poster | A challenge |
|---|
