Semantic Modelling…Is it utopia?

By Suresh G. Nair, Vice President & Chief Architect, Mphasis

Keeping up with constantly evolving regulatory requirements has always been a challenge. Each change requires tracking down the right data, at the right level of granularity, and packaging into the right model. Invariably each need ends up creating a new data silo, with its own infrastructure and scripts.

Data warehouse + data mart infrastructure are one attempt to solve this. However, when we look across these we still see issues. An employee who is also a customer ends up as two separate entities in the system, with very different characteristics. Aggregating these into a cohesive view of the person remains a challenge.

Symptom or problem

A perception we all seem to share is that business changes constantly and it is a complexity that we must live with. Many things in business do change, if we really think about it the entities that we deal with do not change frequently. For example, the characteristics of a customer or an account change in significantly longer cycles than the processes that reference them. If you really think about it, the problem itself did not change much; the changes are in how we solve the problem.

This is much easier to see in other fields. Take transportation. The intention is to move people from a starting point to a destination. People often travel in groups (e.g. families like to travel together). They need to carry luggage with them. They have biological needs that must be addressed. There are physical limitations on what humans can endure.

Given this common problem space, we have hundreds of variations in the solutions – from bicycles, to cars, to airplanes to ships; all the way to space travel. Completely different means of transport, but the “people” aspect of the problem remains largely the same. There are, of course, nuances. Astronauts are much healthier than folks who go on pleasure cruises, in spite of what advertisements would have us believe. These nuances can easily be expressed in terms of what we know as universal truths about people.

Why then, when we model IT systems, do we consider the problem space to be constantly in flux?

Solution is within our grasp

Take a simple example of modelling “people” in an IT system. We know that people have certain common characteristics – names, dates of birth, height, weight, employment status, etc. However, depending on why we are recording the details of a person, some characteristics are relevant while others might not be. It does not mean that the other characteristics do not exist. Instead, to optimize the effort required to capture the details about the person, we only capture as many details as are important for the problem in hand.

The best method, then, would be to model what we know as universal truths, and select the subset that are relevant to each problem.

The most common challenge to this approach is the “boiling the ocean” argument. Are we required to model every single aspect of humans before designing our first application? If you think about it though, it becomes obvious we can build these models incrementally, capturing new “truths” into a single shared model.

Semantic Models

Today semantic web and model driven architecture standards provide open, patent free and implementation agnostic ways for using this approach in application design, development and operation.

Semantic Web standards from the W3C provide a way of expressing concepts and related universal truths that is agnostic of any single system. The specifics of a solution – logic, algorithms, user interfaces, workflows, system interfaces – can be expressed in terms of these universal truths.

The Model Driven Architecture (MDA) standard from the OMG group lays the framework for the automated generation of applications from these models. The model described in semantic format is termed a “Computation independent model”. We add nuances unique to our solution, and we arrive at a “platform independent model” of our solution. We make specific technology choices, and realize the logic and models into our platform, and we get a “platform specific model” of our solution. The MDA standard provides mechanisms for describing the translations that occur between each of these models, allowing for the automation of a huge number of the activities that we today see as inevitable parts of the software development process. Applications built using MDA methods are already showing savings of 40% in the first build of an application, and over 80% on subsequent changes to the same application.

The long slow haul

The kind of knowledge work automation that is revolutionizing other industries has left the IT industry relatively untouched till now. We see semantic modelling and model driven architecture as an inevitable “catch up” for our industry. It does require significant culture change, and is dependent on new “infrastructure” such as universal domain models for key industries. However, as this new infrastructure starts falling into place, the benefits of changing our approach will be so compelling that we feel change will be seen as inevitable.

Semantic modelling is already mainstream in other industries like retail and media. As the financial services industry starts adopting it, eliminating overlaps and siloes will no longer be an unrealizable utopia but a here and now reality.

Suresh G. Nair is Vice President & Chief Architect at Mphasis. He can be reached at suresh.nair@mphasis.com

About Mphasis

Mphasis (an HP Company) enables chosen customers to meet the demands of an evolving market place. Mphasis fuels this by combining superior human capital with cutting edge solutions in hyper-specialized areas. Contact us on www.Mphasis.com

Advertisement

Follow Us:

Sitemap | Privacy | Copyright © © 2017, WSTA®, All Rights Reserved.