Metadata Extraction Tool

The service extracts metadata from scholarly articles across various layouts and languages, particularly focusing on publications from small to mid-sized publishers that historically lacked the resources to maintain detailed metadata.

This issue is prevalent in numerous disciplines, including German social sciences, where missing metadata is common. Additionally, some documents may lack complete metadata because authors did not supply it in an organized form, despite being assigned a Persistent Identifier (PID), such as those found in Zenodo publications.

This AI-powered tool is designed to systematically extract and structure metadata, improving accessibility and organization. Multiple machine learning models are trained using natural language processing (NLP) and computer vision techniques to interpret documents as either text, images, or a hybrid of both. These models undergo rigorous testing and evaluation on challenging datasets to ensure robust performance.

This service is currently under development.