This is part four in a four-part series on knowledge graphs. 

Read - Part I: What is a Knowledge Graph?

Read - Part II: What are Knowledge Graphs Good For? 

Read - Part III: Build a Knowledge Graph in Ten Steps

Over the last few months, we’ve explored what a knowledge graph is, what they’re good for, and how to build one. In this final entry of the series, we will gaze into the future and examine the broader trajectories of the knowledge graph industry. 

We currently stand on the threshold of a new era for knowledge graphs. Once confined to academia and tech giants, the advent of off-the-shelf graph databases in the last decade led to the adoption of knowledge graphs across a wide range of industries. Even with graph databases, however, it still fell to organizations to understand how to build a knowledge graph. If a company possessed a knowledge graph, it was likely that a graph wonk on the tech team had advocated for it and made it work. 

Today, vendors continue to take steps to make the benefits of knowledge graphs more accessible. These approaches generally fall into three categories: the enterprise knowledge graph platform, the knowledge graph as a service, and the embedded knowledge graph. (See Figure 1.)

Figure 1. Trajectory of the Knowledge Graph Market

The Enterprise Knowledge Graph Platform

The enterprise knowledge graph platform is the most direct successor of the previous generations of knowledge graph tools. This category includes offerings from companies such as Stardog, Ontotext, and Neo4J. Though each takes a different technical approach, they all combine a graph database storage layer with features for building, exploring, and analyzing the graph. (See figure 2.)

Figure 2. Enterprise Knowledge Graph Platform

Putting all the features needed to both build and consume a knowledge graph in a single platform reduces the burden on organizations to manage separate tools for each essential function. In addition, these platforms offer low or no-code consumption environments, so non-technical folks can benefit directly from the knowledge graph. I anticipate that mid-sized enterprises will turn to the platform-based approach to build knowledge graphs since their ease of use makes them a smoother entry point for organizations that have limited engineering resources. 

The Knowledge Graph as a Service

Other companies remove the burden of development altogether, simply providing information in the form of a prebuilt, searchable knowledge graph. These knowledge graphs are themselves the product offering. Diffbot is a prominent example of this approach for web data. It provides a massive graph of more than a trillion facts about organizations, products, and news articles sourced from across the web. Rather than constructing their own knowledge graph of these entities, organizations can simply buy access to Diffbot’s. 

Data platform company Tresata operates a freely available knowledge graph (BADaaS) based on leaks of financial information provided by the International Consortium of Investigative Journalists (ICIJ) and the Organized Crime and Corruption Reporting Project (OCCRP), including the Panama Papers and Paradise Papers. They intend this graph as a public service that enables global citizens to seek out and identify financial crime.

In both cases, whether a paid service or free, the labor of constructing the knowledge graph and the data that populates it are provided to the consumer. All a user needs to do is explore or extract the information they desire.

The Embedded Knowledge Graph

The final category is the embedded knowledge graph. In recent years, software vendors have incorporated knowledge graphs into their application engines. Like the knowledge graph as a service, leveraging an embedded knowledge graph requires no expertise in knowledge graph development. In fact, most application users will interact with the knowledge graph without realizing it.

Data.world, a data catalog company, is a great example of this approach. It provides a consumer-grade search experience for finding and cataloging data assets. This application sits on top of a knowledge graph data.world builds automatically from metadata it gathers from customer data sources, but the average user never even knows it’s there. They simply conduct their searches and view and retrieve data through a graphic interface.

Roam Research takes a similar approach to the world of notetaking. Customers use Roam to jot down thoughts, as with any number of other notetaking apps, but Roam captures these notes in a knowledge graph. This underpinning allows users to link their thoughts and record relationships between notes, providing enormous benefit over traditional filing approaches that isolate ideas. But again, users don’t need to know the first thing about knowledge graphs to benefit from them. 

In addition to these two examples, vendors today use knowledge graphs to power many other types of software applications including supply chain management tools, chatbots, and AI search features. Given the natural synergies we explored in the second blog, I expect next-gen customer relationship management systems and content management systems will use knowledge graph engines too.

Wrap-up

As knowledge graphs become mainstream, all three of these approaches will prosper. Although distinct, together they represent a larger trend toward the wider adoption and democratization of knowledge graphs. This growing ubiquity makes it important for business and data leaders to understand the basics of knowledge graphs even if they don’t intend to build one. Hopefully this series has served as an effective introduction. As always, I welcome feedback from readers. Please reach out or comment below to share your experience with or questions about knowledge graphs.