If you’ve come across a knowledge graph recently, then you know what I’m talking about when I say that these small sunbursts of data are worth their weight in information.
A knowledge graph is a single place where you can find all your data and the interlocking relationships behind that data. Google search is enhanced with this technology, and LinkedIn uses knowledge graphs to boost its search, and business and consumer analytics. But the big explosion of these graphs is in enterprise knowledge graphs, and how ordinary companies are benefiting from the power of these mighty configurations.
Every company should have a knowledge graph
Companies are becoming more data-driven every year. Most companies today are set up with data teams or a group of data specialists that are dedicated to driving informed business decisions from the organization’s data.
For your organization, a knowledge graph will create a specific web of knowledge that is unique to your own domain. Imagine a colorful, visual graph that is very unique, very personal, and very full of relevant information. Analysts can use knowledge graphs to:
- Break down all those data silos
- Find information faster
- Make better decisions
- Uncover a whole lot of hidden insights
This will give you a new approach to insights. By piecing together separate data sources, and finding information faster, you can create a bigger picture of the organization, and deliver a different view of the company from the same data. This adds another dimension to your data analytics, and one that will continue to expand as new technologies are added (we will cover those later on in this post).
Value for the entire company
Over the last eight months, knowledge graphs have become part of my everyday life. I’m the person behind the graph, developing the features that pull the data in the right direction that answers the analysts’ questions. As I’ve been working on them, I have seen the high value that these graphs create for many different people in the organization and the endless features that can be built based upon on it.
With an enterprise knowledge graph, analysts’ lives get a whole lot easier. The graph already encapsulates all the relations of a company’s domain. It facilitates generating those reports by having those numbers crunched even before you start. It is easy to segment users and understand usage and trends with a few simple queries.
Another example would be a financial website that needs to provide answers online to customers with high service expectations. These high expectations need to be met, or they may turn to a competitor for their business. Here, a knowledge graph can add a competitive edge by giving a deeper, 360-degree view of the customer. In the knowledge graph, you’re able to link the customer’s personal data (email, phone number, city, state) with their transactional data (financial plans, monthly payment, deals, purchases, etc.). You may also link to actual search terms and find that the customer was looking for a service agent who speaks Spanish. Connect the data for your service agents who speak Spanish, and you can assign this customer to the agent that speaks Spanish, understands the customer’s financial plans, and is located in a time zone that is convenient to all.
That’s a great customer experience — and one that leads directly to orders and revenue.
New technologies will make it easier
Along with the popularity of knowledge graphs comes technologies that will help find and extract important information. For the past eight months, I have been studying the usage information around knowledge graphs, and trying to capture the user’s intent.
The end result is a recommendation system that will help analysts with their queries. This is going to make the analysts work more productive in the following ways:
- Make querying easier: autocomplete will finish the query with the best options. For instance, sales revenue will be completed with sales revenue in 2019.
- Help surface answers: recommendations will surface answers that were never even thought of as the accumulated knowledge from historic searches learning from what others have asked.
- Keep querying: recommendations for the next question will be made available instantly so the next step of analytics is easily accessible and at-the-ready for the analyst. Without stopping, the analyst can go on to the next level of insights.
This is going to elevate the ease-of-use for analysts using knowledge graphs and point the analysts in the right direction with new searches and recommendations on a daily basis.
Graphs are going to become predictive
One thing that excites me about graphs is the rise of machine learning methods applied directly to the graph. Graph Neural Networks (GNN), and graph embeddings are just a few of the hot topics that are going to change the way we use graphs and ultimately help analysts get even more out of this technology.
Unlike standard neural networks, graph neural networks directly operate on the graph structure. A typical GNN application is node classification, being able to predict the label of the node without ground truth. Graph embedding transforms the property graph to a set of vector representations, enabling easy understanding of node similarity and prediction of new connections based on node similarities. By applying this additional layer of inference on the graph, analysts will be able to ask more complex questions, adding an even deeper level to the recommendation solution we talked about in the last paragraph.
This is truly state of the art technology, and it is just around the corner.
Even though knowledge graphs have been around for a long time, only now do enterprise companies have the right infrastructure (both personnel and technology) to implement and analyze the insights these graphs hold.
Knowledge graphs will build the connections and relationships of your company and will reveal insights you never knew were there. I can’t wait to see how the technologies we are working on today will unlock an even greater potential to see deep inside the data. You must start implementing graphs in your organization now so you can reach the full data potential for your organization tomorrow.