Don’t do what I say, do what I mean!

Today, personal assistants do a good job with simple tasks like setting reminders, playing a favorite playlist, even calling your significant other when you direct the personal assistant to “call spouse.” The challenge is to perform more advanced tasks like understanding what we mean, as in ordering pizza for delivery. The concepts of ordering pizza will require the understanding of product choice, payment and delivery. Linking fundamental product concepts to spatial, temporal and payment events requires a knowledge model. Knowledge models can come in all forms and sizes with most vendors implementing them in their own proprietary format. However, there is a substantial drawback to developing knowledge models in a proprietary format, especially when it comes to capturing complex global information, like in the banking industry.

The problem in global context

The banking industry has recognized the risks to business without global information standards. The Basel Committee on Banking Supervision (BCBS) noted in a June 2012 paper on data aggregation and risk reporting: “One of the most significant lessons learned from the global financial crisis that began in 2007 was that banks’ information technology (IT) and data architectures were inadequate to support the broad management of financial risks. Many banks lacked the ability to aggregate risk exposures and concentrations quickly and accurately at the bank group level, across business lines and between legal entities.”

Providing a global solution

The realization that the current approach for architecting complex interconnected banking data is inadequate and requires a new approach to address the broad financial risks.

In response to BCBS-239, the Enterprise Data Management Group (EDM) has developed a Financial Industry Business Ontology (FIBO) based upon Semantic Web open specifications W3C Web Ontology Language (OWL). Developing ontologies (Knowledge Models) based on open specifications answers the problem of building one monolithic data model that cannot handle different types of data, or data that have different meanings, with alternate spellings. Open specification ontologies allow for distributed development with different groups building knowledge models based upon their domain of interest and bridge them all together. As Michael Atkin, Managing Directory of EDM council put it, “forming a harmonization of meaning across data repositories.”

Open specification ontologies not only provide the capability to have precision associated with meaning they also has the capability to perform inferencing. Inferencing is the ability to assert a fact through defining a relationship that is between a subject and an object. From the assertion of facts, inference of additional facts can be obtained.

Delivering on the promise

The below example is a genealogy knowledge model implemented in W3C OWL using an open source Ontology editor called Protégé. The ontology defines five assertions for the 19^th century industrialist Andrew Carnegie. It shows that Andrew has a sibling named Tom Carnegie and a parent, William Carnegie. The model also asserts that Andrew is a male and captures his birth and death date. See Figure 1 below.

Figure 1. Asserted relationships of Andrew Carnegie

To infer new relationships in the ontology you would need to run an inferencing engine, a software program referred to as a reasoner. The reasoner can generate a large amount of pertinent information about the subject from a small number of assertions applied to each individual in the ontology. See Figure 2 below.

Figure 2. Inferred relationships of Andrew Carnegie

All the relationships highlighted in yellow were inferred. Not only did the reasoner identify new relationships, it also identified what roles Andrew Carnegie belongs to based upon the inferred relationships (see box outlined in red).

And if you were wondering if the system was able to infer Andrew Carnegie’s age, see below Figure 3.

Figure 3 – Andrew Carnegie age inferred at death

This is why big data and machine learning will never be able to deliver on the promise of new information discovery. It simply doesn’t understand the concepts of the data that it is operating on.

Leaders in the field

Some technology leaders may consider this modeling approach to be “too academic.” However, over twenty banks and investment firms have adopted this approach and are contributing to the FIBO standard. Some of these financial institutions include Bank of America, Chase, Citibank, State Street, Wells Fargo, Goldman Sachs, Morgan Stanley and UBS to name a few. Additionally, several regulatory agencies have shown interest in FIBO both here in the U.S. and Europe.

Web Ontology Language (OWL) knowledge modeling capabilities are too disruptive to write off as theoretical. Wells Fargo is using these capabilities to define and execute regulatory compliance rules to provide traceability against their banking transactions.

The reality of developing knowledge models isn’t a trivial exercise and will require formal ontology training. However, on the up side, industry experts agree the benefits are substantial; 90 percent decrease in maintenance, 60 percent decrease in operations, and a 30 percent decrease in total cost of ownership.

If you are thinking about dipping your toes into the semantic world, there are a couple of options for getting started. Stanford University, California offers a 3-day intensive training on ontology development, including the use of the Protégé editor and OWL. Additionally, Semantic Arts consulting services also provides week long training on Designing & Building Business Ontologies.

Continuing to use the current data architecture approaches will no longer be viable to stay competitive. The ability to understand what your customer means versus what they say is the competitive advantage that will determine which companies will lead in the 21^st century.