The more we advance in machine learning (ML) and artificial intelligence (AI), the more we realize how exquisitely complex human intelligence is.
“There is no intelligence in artificial intelligence,” said AI expert and colleague, Natalia Modjeska, “or at least not in its current incarnation, which is really just statistics (or mathematics) on steroids. The original AI (a.k.a. artificial general intelligence or ‘good old-fashioned AI’) set out to simulate human intelligence. What’s taking the world by storm today — applied AI or narrow AI — is something very different. It is big data and layers and layers of complex mathematics used to solve very specific and narrowly defined problems. Applied/narrow AI is shallow (it has no knowledge), greedy (it needs big data), and brittle — it does not transfer easily from one application area to another and it breaks down quickly with even the tiniest changes in the data. To get to that original dream, we’d need to perhaps supplement current statistical AI with semantics and world knowledge (and maybe a physical body). Or maybe take it in a completely different direction to avoid the brick wall that the field is heading into.”
This is all to say we have data — a lot of it — and the computing power to help us track virus outbreaks and deliver public services more quickly and accurately than ever before. But the human element, the implicit knowledge we’ve built through experience that we’re not even aware of, is what allows us to make decisions and carry out much of our day-to-day activities.
So how do you make tacit human knowledge into something explicit and repeatable that a machine will understand?
Knowledge graphs allow us to start emulating the implicit functions of the human mind and combine it with the computing power of machines to extract real meaning from our data, whether the data is from within the enterprise — housed in structured data sets and unstructured documents — or from outside in the rich body of community dialogue.
The answer to bridging this gap is in how the knowledge graph:
- Makes connections between lots of different data points, and
- Adds meaning or context.
Connecting the Data Dots
Humans excel at associating their current environment with experiences and facts stored in their memory banks. Families at holiday dinners know the cues that bring on stories about “that summer vacation.” In business, there is no more reassuring and loyalty-building customer service response than, “I’ve been there and here’s what I did.” It’s reaching into a vast pool of experience and surfacing the right answer.
A ‘graph’ represents what information or data is available and how it connects in an organized picture, a flexible model of our knowledge domain. Linking structured and unstructured data provides better quality data for machine learning and artificial intelligence, so that you can build a unified 360-degree view of the customer or gather useful recommendations.
You’re already seeing knowledge graphs at work in Alexa and Siri voice assistant devices and in Google searches. Google’s own version of a Knowledge Graph collates and organizes millions of pieces of data about people, places and facts to create meaningful and accurate interconnected search results.
There’s too much data out there for us to continue to rely on our minds alone to make all these connections.
Related Article: Will Microsoft Graph Deliver on the Promises of the Social Graph?
Adding Meaning to Data
I was a high school science teacher in Europe. Some of my students asked for help writing a card expressing to their teachers how much they would miss them over the break and used the words, “We will yearn for you.” I spent some time explaining why ‘yearning for’ and ‘missing’ someone meant very different things, even though both were listed side by side in their dictionary.
The term semantic refers to the meaning of data. It helps users understand what the data is about and when to use it. We add context to data using an ontology, such as a label and description, that is closer to our own language. Using a more natural language to interact with data means better search results and better co-working between data providers and data users.
Related Article: Enhance Your Customer Experiences With Semantic Data Enrichment
Getting Started With Knowledge Graphs
The knowledge graph is a new layer that lies on top of your existing databases and sets you on a course for artificial intelligence since it can be interpreted by both humans and machines. Below are some tips to help you implement and maximize the value of a knowledge graph:
- Begin with a single use case, linking just a few data sets and reports, and add data and links to it organically so that it’s a dynamic structure.
- Once you have a use case, identify the content you’ll need and classify it according to a taxonomy. While you can refer to industry standard taxonomies for ideas, invest the time to make the taxonomy meaningful for your organization and understand how users organize their information. Buying taxonomies out of the box or contracting a consultant to do it for you is bound to lead to problems.
- The organizing structure becomes even more powerful — an ontology — when you use semantic indexing to replace users’ own words with synonyms to better understand what they mean. The requester doesn’t need to know the exact label to retrieve the information they want.
- Engage business users in the continuous development of the knowledge graph, along with taxonomists, information architects and data scientists.
- Add descriptive metadata to the knowledge graph, such as the version of the report or data lineage, so that users can decide whether it’s the right data and if its quality is acceptable.
Lisa Ehrlinger and Wolfram Woess offer the following definition of a knowledge graph in their research paper: "A knowledge graph acquires and integrates information into an ontology and applies a reasoner to derive new knowledge."
Their definition depicts the essential makeup and power of a knowledge graph: the annotated and validated data, the flexible organizing structures and relationships, machine learning algorithms and human expertise.
As an example, medical researchers are identifying novel therapies by applying knowledge graphs to structured and unstructured data (both public scientific journals and proprietary data sources) to rapidly find correlations between drugs and the diseases they treat. The semantic data integrations allow the researcher to recover valuable data from several public and commercial databases, even if industries use different vocabularies to describe it.
By extracting hidden knowledge from proprietary databases and aligning it with public datasets in a larger context, our results become precise, relevant — and very human.