Making Metadata Meaningful
Metadata is essentially "data about data." It provides descriptive, structural, and contextual information, making other data easier to understand, locate, and use effectively. By capturing essential details—such as a dataset's origin, structure, purpose, relationships, and meaning—metadata enables data to be organised and contextualised.
Currently, much of the focus in AI is on high-profile generative models, representing only the visible tip of the AI Iceberg. However, beneath the surface lies the foundation: data. What sits below determines what can happen above, so organisations must first organise their data by prioritising metadata.
Metadata can be divided into four main types:
🔵 Descriptive Metadata: This includes information that helps identify and locate data. For example, a book’s metadata might contain the title, author, publication date, and genre. In a digital setting, descriptive metadata could include tags, keywords, or descriptions, making data easier to search and retrieve.
🔵 Structural Metadata: This describes how data is organised or formatted. For instance, in a database, it might define table relationships or document structures, helping to ensure data is correctly interpreted, stored, and processed.
🔵 Administrative Metadata: This encompasses information needed to manage data, such as data ownership, access permissions, or retention policies. Administrative metadata is crucial for data governance, ensuring data is properly maintained and protected.
🔵 Semantic Metadata: This connects data to meaning, especially in the context of AI and knowledge management. Using ontologies and knowledge graphs, semantic metadata establishes relationships and contexts, helping AI to "understand" distinctions in data—such as the difference between a "financial asset" and a "physical asset."
Here is a key insight: Semantic metadata gives data meaning. By adding semantics to metadata itself, we make all metadata meaningful.
While this might sound complex, it’s quite achievable in practice. By using ontologies and knowledge graphs, you can unify Descriptive, Structural, and Administrative metadata within a semantic framework. This creates a single Semantic Layer over all your organisation's data.
AI can assist in building this Semantic Layer over your data, leveraging the general semantics of natural language. It can then use that Semantic Layer to interface more seamlessly with the specific semantics of your organisation's data when answering questions at runtime.
The concept is simple, but implementing it requires time and effort—and time is running out. Organisations need to redirect resources from prototype AI projects and vanity showcases to the real task at hand: preparing their data to be effectively utilised by AI.
⭕ The AI Iceberg: https://www.knowledge-graph-guys.com/blog/the-ai-iceberg
⭕ DPROD - A Practical Way To Get Started: https://www.knowledge-graph-guys.com/blog/data-products-ontologies
