Earlier this week, the little journal publishing company I run — the Journal of Bone & Joint Surgery, Inc. — launched a new online presence that includes two new complementary products and a number of new content flows and product entry points.
As part of this redevelopment of JBJS’ online presence, we moved from HighWire Press to Silverchair‘s new SCM6 platform. While my admiration and affection for HighWire stretches back into the mid-1990s and remains unabated, the lure of building on a semantic platform geared for medicine proved irresistible.
The allure of semantics for me is clear — getting the concepts right instead of relying on the vagaries of wording. Authors and editors try to be cute, are fairly undisciplined and unpredictable across a span of years and regimes, and are idiosyncratic by nature, as is the language they use. Semantics can normalize these effects out, make legitimate linkages between concepts, and generate better information faster. Also, if a development concept is clear, semantic content can be redeployed rapidly to create prototypes or full-fledged products.
Over the years, I’ve played with many product or platform add-ons that promise to deliver semantic solutions. They routinely fail to fulfill their promise. Someday, they may. OCR took a long time before it hit primetime, as did voice recognition. But I didn’t see any sense in waiting for an extramural (or “bolt on”) solution when a carefully crafted, integrally semantic platform existed in the precise space (medical publishing) I work. As one savvy VC person told me a few years ago, medicine has a huge advantage over nearly any other field — it has a stable and carefully curated lexicon. Someone captures that in a platform? Yeah, I’m there.
While the integrations we’ve been able to achieve across products and against other semantic products like Guidelines.gov are exciting and useful, there’s more here than just smoother content connections. One of the more exciting products we’ve just launched is JBJS Case Connector, which takes case reports from 1900 to the present — more than 2,500 of them — and semantically interlinks them. Already, I’ve been able to surprise one editor who thought a concept flagged in an article couldn’t possibly be in it, at least judging by the title. But as we delved in, the concept was clearly there, and significantly. In fact, it was the core concept of the work. Yet, judging from the headline, you never would have guessed. It was a beautiful illustration of how writers and editors can work against themselves, while semantics aren’t so easily deceived.
Competition is good for everyone. It’s salutary and enlivening. The emphasis on semantics that is sweeping our industry is also good. Competition on semantics is already underway, at least in the marketplace of ideas. We should be trying to generate knowledge, connections, and clarity. Semantics provide a promising path forward. Some solutions will be better than others. Not all semantics are created equal. (Oddly, the semantics of “semantics” requires some careful reflection.)
Semantics have long been seen as the future of content in the networked world. For JBJS, they are now part of our present. And I’m personally very excited about what lies ahead.
Medicine has another huge advantage over other fields when it comes to developing semantic technologies, namely Federal funding. The National Library of Medicine has an annual budget of about $350 million and has poured many millions into semantic R&D over the last two decades.
In the physical sciences the closest thing to NLM is the Energy Department’s Office of Scientific and Technical Information. Even though DOE funds 40% of all the Federal basic research in the physical and energy sciences, OSTI’s budget is just $9 million. So there really is no NLM for the physical and energy sciences.
On the other hand, this makes OSTI something of a skunk-works. Our semantic tools are really simple compared to NLM’s, yet they are powerful in their own way. For example, I just developed an algorithm for finding all-and-only the core papers on a given topic. It doesn’t use any of the elaborate semantic machinery that NLM thrives on. No RDF triples, ontologies, thesauri, tagging, etc., so it is completely portable.
The point is that there are a lot of different semantic approaches and technologies, with different price tags, and this is not well understood. Moreover, different fields are in very different places when it comes to semantic tools, many of which are not portable from one field to another.
David Wojick, Ph.D.
Senior Consultant for Innovation
DOE OSTI
Posted by David Wojick | Jun 1, 2011, 7:27 am