The Benefits of Learning

What 'All the Data' Really Means

Kendall Clark
Jul 25, 2017 · 4 min read

We’ve integrated machine learning in Stardog 5. In this piece I explain how that benefits our users.

TLDR: Machine Learning with access to all the data means better insight, faster. If you want to know how machine learning (ML) works in Stardog 5, read Learning to Predict. But if you want to know why we integrated ML into Stardog then read on.

What is Stardog?

The Stardog Knowledge Graph is a full-service stack devoted to solving the Enterprise Data Silo problem. A Knowledge Graph is a Knowledge Toolkit deeply integrated with a Graph Database. Machine Learning is an important part of Stardog’s Knowledge Toolkit.

Thomas Eakins, The Chess Players

Learning the Enterprise Knowledge Graph

But how specifically does ML benefit Stardog’s customers and users?

To answer that question let’s consider the two main types of AI-enabled startup:

  1. generic AI capabilities and services “in the cloud”
  2. full-stack solutions to strategic business problems that use AI as an enabler

In short, we think that (1) are all doomed and (2) is where the wins will happen. Type 1 startups have no economic moat and no chance to build one. Type 2 startups do.

Stardog is a Type 2 startup. We are a full-service stack and the strategic business problem we’re devoted to solving is the problem of enterprise data silos. And machine learning will help us get there faster.

All enterprise IT — and, in fact, most all of enterprise period — will in the future run on data; and that will enable Type 2 startups to thrive. Stardog is purpose-built to enable all enterprise activities, including other startups, to more fully monetize data as the strategic asset by solving the enterprise data silo problem.

How will putting ML into Stardog help our customers do that? Two ways: building the knowledge graph; and deriving actionable insight from it.

Jean-Louis David, The Death of Socrates

How to build a Knowledge Graph

The three biggest knowledge graphs — Google, Facebook, LinkedIn — on the planet are built with, in part, machine learning (ML) techniques. Why? Because these knowledge graphs are made from other people’s data — the web itself, personal life, and professional life, respectively. Those people and orgs, at least in principle, own their data and can change, republish, improve, degrade, truncate, or amplify their data whenever and as often as they want.

So Google-Facebook-LinkedIn use automated measures to respond appropriately to that unending and non-negotiable flow of changes. The reason knowledge engineering done by really smart people with good tools isn’t the primary approach any more is that the flow of data is too fast, too large, and too diverse. Knowledge Graph creation is (partially) automated because there is no other way.

Better Insight Now

While the data landscape “behind the firewall” is quite different in some respects from what Google faces on the public Web, there is also a lot of overlap. We know that many of the ML techniques that the Big 3 employ work in the enterprise, with suitable modifications.

So ML helps us get to the Enterprise Knowledge Graph faster. And then that superior data accessibility creates a virtuous cycle between greater data unification and better actionable insight based on what the organization in sum knows. The thing about an enterprise knowledge graph is that it should know stuff.

Knowing stuff is a lot like learning stuff. People who know and learn stuff are good. Machines that know and learn stuff are good, too. The combination of people and machines are the best in part because the combination of better (i.e., all) the data and the right algorithm is the best thing of all.

Reubens & Brueghel, The Feast of Achelous

What’s next

Knowledge graphs know stuff; your enterprise knowledge graph should know stuff about your enterprise. And it should learn more stuff from the stuff that it already knows. That’s where we’re headed. Stay tuned for more ML and more AI in Stardog:

  • graph extraction from dark data
  • probabilistic inference over graphs
  • knowledge graph construction, including data cleansing and quality
  • structure learning, including refining schema alignments
  • analytics beyond predictions

Download Stardog today to start your free 30-day evaluation.


Originally published at stardog.com.

Kendall Clark

Written by

CEO & Cofounder of Stardog, the world's leading Knowledge Graph for the Enterprise.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade