2024 Volume 8 Issue 3 Pages 111-114
The preparation of unified metadata is essential for facilitating federated searches and the reuse of data across organizations and fields. The distribution platforms for data catalogs are decentralized, but the quality of metadata, which is crucial for integration, is low, and there is a shortage of data curators. If metadata for federated searches can be automatically generated from basic attributes that are readily available, such as descriptions and titles, it could address these challenges. This report introduces two studies on metadata generation using generative AI, published by the authors in 2023. It highlights the potential and challenges of basic information extraction using LLMs and metadata completion using a combination of LLMs and knowledge graphs.
