The task of harmonizing disparate applications has been around for years, but semantic interoperability aims to make the job easier. The goal is to eliminate the language bottlenecks that arise when systems that were never intended to talk to each other attempt to do so.
Those barriers arise when one term has multiple meanings or two or more terms refer to the same thing. A search query that generates too many or too few responses is one familiar consequence of a semantic breakdown.
Human intervention can hammer out differences in meaning. But semantic interoperability would have machines handling those negotiations.
Recent moves to commercialize semantic technologies have increased interest in the topic. At least two broad-based projects specifically target health care: the World Wide Web Consortium’s (W3C) Semantic Web Health Care and Life Sciences Interest Group and the federally oriented Health Information Technology Ontology Project (HITOP). Both were launched in 2005.
Those groups have spent the past few months raising awareness about semantic interoperability and its health care implications.
“The challenge has been bringing the health care field to a level of awareness regarding the need for semantic interoperability,” said Marc Wine, chairman of HITOP. But he said the field now understands that it “should become an integral part of the mission for making electronic communication in health care accurate, efficient, reliable and secure.”
“It has enormous potential,” said Robert Coyne, executive partner at TopQuadrant, a company that makes tools to support semantic interoperability projects.
The semantic challenge
An essential problem with data sharing stems from every system having its own way of representing data. Relational databases, for example, each have their own schema for defining tables and fields.
“It’s very difficult to share data in relational databases,” said Susie Stephens, principal product manager for life sciences at Oracle. “It’s hard to merge relational schema and hard to understand someone else’s schema.”
Even Extensible Markup Language (XML), a technology designed to ease the exchange of data, has limitations, Stephens said. “The semantics aren’t explicit within XML,” she said. XML imposes a certain grammar, or syntax, but machines may still stumble on semantics.
For example, a physician knows that dropsy and congestive heart failure could refer to the same ailment, said Charles Mead, a senior associate at Booz Allen Hamilton. But a computer wouldn’t know if the terms are similar or different.
Mead cited another example: A common lab test — serum sodium — may be represented by serum NA or serum NA++. It may also be embedded within a larger test under a different name.
A hospital’s laboratory system, however, will recognize only one of those variations.
It’s those kinds of problems that semantic interoperability seeks to address.
“Semantic interoperability is trying to bring together the meaning of data in multiple systems in a way you can pull it together and make use of it for an application,” said Les Westberg, senior software architect and engineer at Northrop Grumman IT.
Westberg said he views semantic interoperability as focusing on the interoperability of systems within a certain domain. He differentiates semantic interoperability from the Semantic Web initiative, the goal of which is to extend the concept to the entire World Wide Web.
The Semantic Web aims to achieve a “common framework that allows data to be shared and reused across application, enterprise and community boundaries,” according to W3C, the group the leads the Semantic Web effort with help from academia and industry.
The Semantic Web’s global scope makes the task much more difficult than smaller-scale semantic interoperability, Westberg said. “If I’m trying to get semantic interoperability…across 20 systems, I know something about those 20 systems,” he said. “On the Web, I don’t know what I’ve got out there.”