Why Inference Is Key to Realizing Data’s Full Potential

Leveraging a knowledge graph’s inference capabilities, organizations can extrapolate new data connections and explain any new connection they create.

Digital transformation is all the rage, and in most cases, the goal of digital transformation is to treat data like an asset. In some instances, that means monetizing data, and in others, the goal is to leverage data more efficiently to derive insight to make better decisions. However, in reality, both are hard to achieve. Digital transformation demands rapid insights from increasingly hybrid, varied, and changing data, but traditional data integration platforms were not designed for today’s environment. As a result, organizations can no longer keep up with data’s growing complexity, nor can they identify hidden relationships and connections between the data to uncover new opportunities. Increasingly, what’s needed is an inference capability that allows companies to bring different datasets together and analyze them to derive insights.

Driving
growth and innovation in today’s complex world of never-ending data capture is
only possible when IT teams can break free of rigid data structures and outdated
integration styles.

Agility is
key to business success, and enterprises are desperate to make data usable when
it counts, not for answers tomorrow or next week, but for right now.

But deriving
untapped value requires the ability to connect data based on its business
meaning, irrespective of the format, source, or underlying tech. The sheer
amount of data derived from machine learning and other sources require the
ability to associate related information stored in disparate sources and then
apply a rich web of relationships to discover new associations. This is key to
realizing the promise of digital transformation. But how does one go about
achieving this?

Moving from premises to logical consequences: How data
fabrics provide needed inference

Organizations today are adopting modern
integration approaches like data fabrics to power collaborative,
cross-functional projects and products and escape reactive workflows. Weaving
together data from internal silos and external sources, they create a network
of information to power the business’ applications, Al, and analytics. Quite
simply, they support the full breadth of today’s complex enterprise by creating
connections between information stored in disparate sources.

Knowledge graphs are an integral part of an effective data fabric as they create a reusable network of information, represent data of various structures, and support multiple schemas. Creating the semantic understanding of enterprise and third-party data, knowledge graphs serve as the core of the data fabric– enriching and accelerating existing investments and providing critical access to business insight. More importantly, knowledge graphs turn data into machine-understandable, real-world knowledge that supports situational changes, so meaning alters depending on circumstances. Once established, the knowledge graph also uses this rich web of relationships to discover new associations within the data. These inferred relationships create a richer, more accurate view of an enterprise’s data.

By providing
layered associations between concepts, knowledge graphs provide nuanced
understanding so knowledge-driven organizations can identify new discoveries.
They also provide the context that is often missing from data because the
knowledge graph is purpose-built to support the fluctuating nature of
knowledge. The result is a more flexible foundation for digital operations as
the technology easily accepts new data, definitions, and requirements.

The
knowledge graph’s data model, often called an ontology or vocabulary, lays out
common relationships between entities and allows companies to describe complex
domains. Consider medicine as an example. To develop a new therapy,
pharmaceutical companies must have access to multiple facts, modeling
constructs, and business rules, all of which must interact with each other to
imply new connections. This inference capability is what makes it possible for
manufacturers to link people to infrastructure via the applications they use.
It also helps them to apply controls based on the similarity of new incidents
to past incidents and find inferred links between investigators and therapeutic
areas based on the conditions being investigated in studies. And the list goes
on.

Applying multiple data models to a data fabric at the same time enables organizations to support multiple applications that require different interpretations of the same data. Traditional data integration approaches such as data lakes or data warehouses are limited in this capacity, as they make it difficult to support more than one schema. This is one reason enterprises have to continually create new data silos for each new application, project, or analysis. Such an approach reduces the ability to perform inference analysis.

Enabling the
connected enterprise-additional components of a successful data fabric

Leveraging a
knowledge graph’s inference capabilities, organizations not only extrapolate
new data connections but also explain any new connection it creates. In
contrast to black box recommendation systems, which cannot provide any
explanation or rationale for their results, the knowledge graph can explain all
inferences and results in terms of data, schema, and business rules. This
explanatory transparency enables users to review how the knowledge graph arrived
at an answer and the business logic referenced to do so. This is not only
critical for providing trusted results and accountability within an
organization but is also necessary for certain legal and regulatory
requirements.

While a
knowledge graph is the key ingredient of the data fabric, it is not the only
thing an organization needs to be successful. An effective data fabric requires
leveraging and connecting existing source systems. It also requires the ability
to connect to existing data catalogs, data lakes, databases, and other data
management platforms. For data fabric deployments, leveraging work completed in
data catalogs is key to accelerating data discovery and semantic enrichment.
Using the data catalog as an input, the knowledge graph builds a data map of an
enterprise’s data assets which further accelerates data fabric creation through
partially automated learning and auto-mapping of existing sources.

Creating an
enterprise-wide data model is another common question regarding deploying a
data fabric. Many think this is a potentially expensive and time-consuming
prerequisite to the initiative, but, in reality, they only need to define as
many concepts as needed for their initial use case. Start by identifying a
critical business problem to spearhead the broader data fabric initiative.
Approach the data fabric with an MVP mindset and focus only on the minimal
amount of work needed to accomplish the first business objective.

Organizations
of all sizes are placing an even greater focus and investment on digital
transformation. Despite this renewed attention, fundamental data challenges
remain a primary obstacle. Digital
transformation requires data mastery and thanks to the legacy of IT is not a
simple thing to achieve. There are just too many things to manage: data
formats, standards, data types, velocities, schemas, systems, databases, silos,
methodologies, models, etc. The sheer diversity of the modern enterprise IT
landscape is daunting.

Leveraging the modern approach of
knowledge graphs, organizations can not only connect their internal data silos
in a meaningful new way, but they can also discover hidden facts and
relationships through inferences that would otherwise be unable to catch on a
large scale. By capturing the nuanced meaning that different business units may
have for the same entity, organizations can create a reusable digital foundation that keeps
pace with continued shifts in the market and be prepared for whatever comes
next.

This UrIoTNews article is syndicated fromRTInsight