Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 25·5 min read
Automated Construction of FAIR Digital Object Knowledge Graphs from Flat Cultural Heritage Records
TL;DR
A pipeline was developed to convert flat Europeana cultural heritage records into FAIR Digital Object (FDO) knowledge graphs, achieving 86% metadata linkage. This automated process enhances the machine-actionability of cultural data by ensuring every entity is resolvable via persistent identifiers.
✦ Why It Matters
Implement this pipeline to enhance the machine-actionability of your cultural heritage data today.
Key Takeaways
- Utilize the pipeline to convert flat records into FDO-compliant graphs.
- Leverage large language models for metadata classification and linking.
- Adopt controlled vocabularies to enhance data interoperability.
- Evaluate cross-lingual term merging for improved data accuracy.
How It Works
The pipeline automates the classification of metadata values using a large language model, which determines whether values should be converted into persistent identifiers or remain as literals. It routes these values to controlled vocabularies, ensuring that each heritage entity is represented as a distinct FDO with its own PID, enhancing the graph's machine-actionability.