How To Start Solving Data Challenges With Knowledge Graphs

In 1991, Sir Tim Berners-Lee launched the first ever website. Although simple, it represented years of research into how best to share documents within networked environments. The web quickly evolved out of a purely academic pursuit to be the backbone on which so much of the modern world is built. But the web had a flaw. Computers can’t understand the contents of web pages the same way that human brains can. This is still the case despite recent advances in NLP, computer vision and machine learning technology.

In the late 1990s, Lee proposed the Semantic Web — or, adding a layer of metadata to sites, giving the concepts represented by web pages machine-readable definitions and hierarchies. The Semantic Web hasn’t achieved widespread adoption, although development continued, leading to the popularization of knowledge graphs in the early 2010s.

Knowledge graphs built upon ontological concepts developed as part of the Semantic Web but allowed for a more expressive hierarchical system, with a greater emphasis on scale and relationships. This made it well-suited for AI-driven tasks and fueled its adoption within Google, Apple, Uber and others.

In the years since, interest in knowledge graphs has continued to grow. If you’re contemplating adopting them within your organization, this article will help you get started, from recruitment and planning to building your technology stack.

Knowledge graphs are well-suited to organizations with large data sets and where extracting knowledge often proves burdensome. For example, an organization might use a variety of data and content management systems, all without the ability to communicate with each other. Data may be structured in a way that’s mutually incompatible. Or, the issue may be cultural, with little horizontal collaboration between teams and departments.

Your pain point will be distinct to your organization, so too will your application of knowledge graphs. Some financial services organizations use the technology to weed out potential costly compliance and fraud issues, while some in the biosciences sector focus on its potential to expedite drug discovery and development.

The objective will be the same: taking a collection of messy, unstructured and scattered data sources and unifying them and integrating them with the knowledge that gives data meaning, so they can be better analyzed and provide more immediate value.

When trying to build out your knowledge graph team, you’ll find yourself struggling to identify suitable candidates with firsthand experience. Knowledge graphs are a relatively recent innovation, and your organization will be forced to compete with the deep pockets of Uber, Facebook and Google for talent.

Look for adjacent skill sets. Experience with the standards in the Semantic Web and knowledge representation space (like RDF, OWL and SPARQL) or graph databases (like Neo4j and Amazon Neptune) is helpful. Failing that, look for a strong track record for building and managing data systems.

Interestingly, a new role has grown out of this trend — the knowledge scientist.