The Underlay is a project to make machine-readable public knowledge available to all as a public good.
The creation of richer collections of machine-readable knowledge is inevitable. What is not inevitable is that these knowledge bases will be a public resource, freely available to all. In fact, many large corporations are now engaged in redundant efforts to build their own private knowledge graphs for their own needs. We believe everyone, including these companies, would benefit from making public data more easily accessible to all.
Without the Underlay, only those that have the resource to gather and maintain data will be able to take full advantage of it. With the Underlay, public knowledge will be cooperatively gathered and openly shared as a public good.
Our knowledge of the world is dynamic, changing with context and over time. As both knowledge production and disinformation accelerate, it is increasingly hard to assess the reliability of information at first glance. We need new, public frameworks for understanding the context and sources behind the data and claims that power our world.
The Underlay is the structured knowledge that lets us talk about the relationships and interactions of our world. If the internet is about giving you access to a book about Marie Curie, the Underlay is about giving you access to information that helps you understand Marie Curie: the cities where she lived, her authored papers, each paper’s co-authors, those co-authors' cities of residence, the location of Curie commemorative statues around the world, and so on. If the internet is about content, the Underlay is about context.
The Underlay represents knowledge in the form of relationships between entities, for example the relationship between an author and a book, or a location and its current temperature. This method of representing knowledge is called a knowledge graph, and it can be used to represent any type of machine-readable information.
The Underlay differs from most other knowledge graphs in that it represents not only the information itself, but where it came from and what others have to say about it. Since what is trustworthy and relevant depends on the application, information in the Underlay is not assumed to be true. Instead, it is designed to be curated by others into databases that contain only the information that they judge to be trusted and relevant for their specific needs.
The Underlay draws its content from public information sources, including public records, scientific publications, commercial catalogs, the web, and open databases created for the public good. The Underlay also provides a convenient way for for-profit, not-for-profit, and government organizations to publish machine-readable information that they wish to make easily available, including schedules, catalogs, scientific datasets, public disclosures, standards, and more.
The Underlay is not a single piece of software or a single database, but rather a suite of interfaces to connect, share, align and query machine-readable knowledge about the world from different sources.
Elements of the Underlay include:
Languages and protocols for describing assertions and their provenance, and compiling them into curated packages
Programming libraries for integrating Underlay data into applications and workflows, or overlays
The Underlay does not replace other databases, it makes it easier to combine, link, and reference them, keep them up to date, and share them with others.
Like the Internet, the Underlay is an idea whose implementation will involve protocols, agreements, computing power, and cultural norms. The project is built, funded, and maintained by an open Consortium of knowledge providers, archives, and infrastructure developers; universities, non-profit organizations, and companies.
The Underlay Consortium is developing software to enable members to choose which parts of the Underlay they will store and maintain. These registries will keep one another automatically updated, and allow members to maintain data based on their own criteria of trust, utility and relevance. Members are free to use and share the data in any way they want; we expect them to serve either a specific constituency or the public at large.
If you’re interested in joining the Consortium, write to email@example.com.