Knowledge graphs may not be as lauded as machine learning, as well known as Natural Language Processing, or as futuristic as their synthesis in conversational AI applications, but they’re an equally vital—if not necessary— component in the modern cognitive computing stack.
In fact, these semantic graph repositories oftentimes provide the foundation upon which machine learning and NLP rely on to deliver practical business value across organizations.
The two pivotal aspects of knowledge graphs that make them ideal for statistical AI deployments are their naturally evolving, mutable data models and standardized vocabularies, which extend into well defined taxonomies.
Consequently, semantic graphs are ideal for modeling various data types, structures, and formats because they harmonize them to provide highly contextualized relationship discernment between even the most disparate of datasets.
One of the most pragmatic use cases for this technology is to standardize data across the enterprise for cross-department access to the most relevant information—and its underlying concepts—within an organization.
According to Lore IO CEO Digvijay Lamba, “Getting information across the business is super important, and this is true for almost any industry. In any industry the sales guy, when he goes and talks to prospects, he wants to know did the marketing [team] send an email? Are there any previous contracts? You need all that information in order to work with a customer.”
Knowledge graphs store data as facts in the form of declarative statements. Aside from their intelligent inferences that derive new facts by reasoning about existing ones (which is critical for understanding relationships in datasets), their primary cognitive computing benefit is their capacity to standardize data despite inherent differences in data sources.
This capability enables organizations to leverage common data models across the enterprise for any use case. The underlying data model expands to include various business concepts pertinent to respective departments, because they’re all founded on the same model. “Everything a business team is doing will kind of be related to what the master data model is, or what other teams are doing,” Lamba disclosed.
The utility of this approach is tremendous. Different departments can tailor the model to their own needs in terms of business concepts, their terminology, and their tools of choice. Collectively, however, the consistency of this unified model spans across departments “so you’ll be able to look at the differences, see how they’re related to each other, and govern that whole model in some sense,” Lamba acknowledged. Without this unified approach, organizations are simply creating data silos that are costly to integrate across the enterprise.
Recurring Business Value
Due to the harmonization of semantic graph data models, a salesperson can readily determine any previous interactions (across business units) a company had with prospects to get relevant information about marketing and pricing, for example. This insight will almost surely help achieve better business outcomes. A horizontal view of data across the organization behooves customer service agents when interacting with clients; this method is also useful as a means of developing comprehensive, 360 degree views of customers.
Due to the centralized means by which semantic knowledge graphs manage the terminology for the concepts represented in their data models, “in the data model you can have dependencies, relationships, etc.,” Lamba commented. “It’s a graph, internally, but you can drive off of other people’s work because of relationships and things like that in the model.”
This characteristic is no mere point of academic interest to data modelers or IT units trying to maintain models. On the contrary, it denotes the fact that business users can employ one another’s modeling concepts as the launching point for their own, effectively enabling them to profit from the efforts of their co-workers. In industries like pharmaceuticals, life sciences, and healthcare this capacity is invaluable for decreasing time to market for new drugs or devising potential cures for harmful conditions.
“You can layer the business logic where what other teams have done, you can leverage the language they have developed and create your own derived language on top of that,” Lamba posited. That language relates to underlying business concepts and whatever research or work went into creating them. Subsequently, different users in respective departments don’t have to recreate valuable work others have done to apply it to their own use cases.
Knowledge graphs’ common data models propagate enterprise knowledge for any particular field, discipline, or use case. Whether that’s to drive sales, perfect customer interactions, or discover new treatments to improve society, this pillar of AI is critical to capitalizing on data-driven processes. When paired with elements of machine learning to automatically map incoming data sources to models and NLP to declare business rules in natural language, semantic knowledge graphs make AI a meaningful enabler of all organizational objectives.
Featured Image: NeedPix
Jelani Harper is an editorial consultant servicing the information technology market. He specializes in data-driven applications focused on semantic technologies, data governance, and analytics.