top of page
  • Writer's pictureShawna Applequist

The Importance of Good Metadata

Since Microsoft launched Project Cortex in November 2019, knowledge experts globally have begun to leverage information through the application of AI techniques and Graph APIs. Knowledge management needs to be embedded into the everyday processes of companies.

Back in 2020, Jed Cawthorne wrote an article titled "The Secret Sauce Behind Project Cortex: Good Metadata" in which he discusses the importance of solid metadata. Cawthorne discusses the features and capabilities of Microsoft Project Cortex indicating that with the introduction of "Topics," automatic creation of topic pages will be embedded into the interfaces of existing Office 365 applications.

"Sounds pretty sexy, no? AI! Machine learning! Azure Cognitive Services! But the foundation of all this magic is something much more humble: metadata."*

Metadata is the backbone of many companies. While NLP and machine learning are becoming useful tools for managing and processing this metadata, humans still play a major role in this process. It takes a human touch to tweak and manage the metadata to ensure it produces quality results.

One of the more important aspects of metadata management is the creation of taxonomies. Using a taxonomy for metadata terms will speed up the time it takes a human to do those manual adjustments and training the artificial intelligence. Taxonomies help these AI engines correlate pieces of data from one document to other terms in the aforementioned taxonomy.

It is exciting to see the advances in machine learning and artificial intelligence. The degree to which these services are evolving is impressive. But, at least for the time being, these services still require an amount of human touch to get the most effective data. And sometimes this process can be extremely time consuming, pulling knowledge workers away from their other tasks in order to train the AI. Taxonomies are just the thing knowledge workers need to speed up time to delivery.


bottom of page