The increasing realization that deep learning alone cannot be the solution to build robust, reliable artificial intelligence (AI) systems, coupled with the ever-increasing need to make use of heterogeneous data sources for decision making, has led to a recent resurgence of knowledge graphs (KGs). KGs are essentially graph-based representations of information that consist of three simple elements: nodes (which represent entities), edges (which encode a relationship between entities), and attributes that describe the relationships and entities. With this simple recipe, we can model any real-world problem as accurately as possible and thus encode domain knowledge into a system that is transparent for humans but can also be interpreted by computers.
KGs have been around for a while (research on them began in the 1980s, and Google announced it was using them in 2012), but they have often been solely used for knowledge representation. Today, even small companies have an amazing amount of data (often heterogeneous), and KGs are the perfect tool to leverage that data. Additionally, various technology platforms and open source tools now exist that make it much easier to design, build, and deploy KGs.
Use cases for KGs vary in range and cross many industries. Their most prominent applications are in product or content recommendation systems, but they have been successfully used in drug discovery research, for the estimation of passenger flows in transport hubs, and in the optimization of global supply chains. These are just a few examples; several others are included in the first article of this issue, authored by Lila Rajabion. Business leaders are discovering that KGs can provide meaningful insights into internal data, empower employees by serving up the right information at exactly the right time, and help managers and others make better decisions.
However, the most exciting KG area relates to AI. As discussed in the May 2021 issue of Amplify (and by Cigdem Gurgur in this issue), the lack of explainability (especially in deep learning systems) is a major challenge for more widespread adoption. In the May 2021 issue, Cutter Expert Claude Baudoin and Clayton Pummill told us:
AI is mysterious. The vast majority of society does not understand how it works, and deep neural networks in particular can produce results that we cannot readily explain. People generally fear what they don’t understand.1
KGs are now playing a seminal role in the emergent field of neuro-symbolic AI, which aims to integrate domain knowledge into AI systems. By combining AI’s statistical/machine learning (ML) side with KGs, we get more effective, more explainable cognitive results and begin creating logic-based systems that get better with each application.2 In other words, we can build the next generation of AI models, ones that support better human-machine collaboration, an idea taken to its very edge by Andy Williams in this issue with his article on general collective intelligence (GCI) and Industry 5.0.
In This Issue
Our first article looks at a number of use cases for KGs, both general and specific. Rajabion provides four examples of how KGs can help leaders advance their understanding of the business environment in which their company sits. These include merging data silos to create a company overview across divisions, connecting different types of data in meaningful ways, aiding informed decision making by narrowing searches and contextualizing information, and showing interconnections that help leaders gain perspective. Next, Rajabion dives into how Google, LinkedIn, eBay, and IBM are using KGs and explains how other companies could follow suit. She then addresses four challenges currently faced by companies looking to leverage KGs, followed by a look at specific business efficiencies enabled by KGs, including making data more accessible for employees, helping leaders make data-driven decisions, and assisting companies in deploying AI technology.
Next, Gurgur looks at KGs in the context of blockchain. The article begins with background information on how KGs have been used in advanced analytics and their role in helping AI developers. Gurgur then shows how blockchain’s immutability and verifiability offer designers a way to advance KGs to produce more reliable results. The blockchain/KG combination is an ideal one to build more explainable AI systems, she says. Finally, Gurgur explains how KG-enabled information systems can be used in industrial settings to enhance product development lifecycles, improve factory safety, and enhance information systems to the point where employees need less technical knowledge to perform their duties.
Our third article is from George Hurlburt, who details how a KG was used to assist a regional center of a major university system in its course selection process. The KG helped leaders more clearly see the array of educational pathways from K-12 to community college (CC) coursework that are the results of articulation agreements between universities and CCs. Hurlburt shares five figures from the KG that demonstrate its meaningful visualizations. He also explains how the KG was built, including limiting the number of arcs and emphasizing node unambiguity. Finally, Hurlburt concludes with five key academic relationships and trends that are clearly demonstrated by the regional center’s KG.
Our fourth article, by Williams, looks at how human-centric functional modeling (a way to allow computers to solve general problems) could be used to create KGs capable of providing compete semantic models of systems, enabling us to transition to Industry 5.0. He defines Industry 5.0 as a world in which far greater integration is possible, including functional computing approaches like GCI. Although the emergence of GCI isn’t guaranteed (it could end up in a technology gravity well, says Williams), it would bridge type 1 and type 2 reasoning and lead to a radical increase in our ability to solve every problem.
Our final article — written by myself and my colleagues at Arthur D. Little, Philippe Monnot and Armand Rotaru — demonstrates KG use in the real world. We illustrate several prominent, real-world KG applications, then detail how we designed a KG to ensure vertical traceability within a systems engineering context. We began by extracting relevant entities from 20,000 heterogeneous files with the help of natural language processing (NLP) technologies and proceeded to define a suitable ontology that incorporated concepts from the field of systems engineering. We then developed an ML model that consumed features derived from the KG and mimicked the way an independent safety assessment auditor would work in practice.
Using precision and recall to evaluate the model’s accuracy resulted in finding previously incorrectly labeled software requirement specifications. We also found that combining graph-based features with text-based ones boosts the classification accuracy significantly, thus showing significant promise in augmenting human safety assessors in the future. We end the article with some specific advice on using KGs, including unlocking new insights, extracting more from the data you have, and starting small with the intention of scaling quickly.
We hope you enjoy reading this issue (and viewing Amplify’s brand-new design); we certainly enjoyed putting it together. We’re hoping KGs’ potential to take important processes and technologies to new levels will help business leaders better connect the dots.
1 Baudoin, Claude, and Clayton Pummill. “Bridging the AI Trust Gap.” Cutter Business Technology Journal (renamed Amplify), Vol. 34, No. 5, 2021.
2 Aasman, Jans. “Neuro-Symbolic AI: The Peak of Artificial Intelligence.” AiThority, 16 November 2021.