Mankind has been trying to understand the nature of time since, well, since forever. How time works is a big question, with many different facets being explored by scientists, philosophers, even social-psychologists. Semantic technologists, however, are focusing a little more strategically, considering temporal data management for semantic data.
At the Semantic Technology and Business Conference in NYC, coming up in early October, Dean Allemang, principal consultant at Working Ontologist LLC will be hosting a panel on the topic of managing time in Linked Data. Relational database systems long have been tuned into dealing with bi-temporal data, which changes over two dimensions of time independently – that is, valid (real world) and transactional (database) time. Not so with RDF databases. But many institutions, in fields ranging from finance to health care, have no desire to go back.
“They’ll lose all the RDF powers they’re familiar with, all the semantic linkages,” says Allemang. “And if you want that kind of distributed data understood in your enterprise, a relational solution isn’t going to help.”
There’s good reason for improving bi-temporal RDF data management. In many instances, transactional data isn’t updated as soon as there’s a change to valid data. “It may be out of date for quite awhile,” says Allemang.
The panel seeks to provide a cross-industry perspective on temporal representation in a semantic setting. Financial institutions, for example, “may take actions based on faulty data, which they may later correct. For any number of auditing reasons, you have to be able to go back to the history of what you knew” at the time that you took those actions, he notes. Indeed, Allemang points out that the Financial Industry Business Ontology (FIBO) community is contemplating questions of how to – or even whether – the standard for defining financial industry terms, definitions and synonyms using semantic web principals and OMG modeling standards should in some way account for temporal relationships.
In health care, the same problem of updates to a database system lagging real world changes – and not knowing just how long the database was out of synch with them – could plague scientific research or personal diagnosis. “When do you come to know that a specific organism has infected a patient? You may not have known until long after the infection happened [what exactly it was], and you took a lot of actions in between,” says Allemang, whose results you may want to trace back to your original, incorrect diagnosis.
The solution to such scenarios? “You do multiple time stamps on the information in your database,” Allemang says. “What you really want at the end of the day is a time machine.” That is, you want to be able to know what you believed to be the case – for instance, that Party A owned Security X – at a specific point in the past when you made a certain decision, even if the actual owner at that point had transitioned to Party B. It’s not about ever assuming that you have the right information – it’s just about saying what you knew as of today or as of yesterday, and marking it up in some way, he explains.
How best to achieve those ends will be discussed at the panel – should industry standards like FIBO mandate details; should businesses pursue their own approaches to incorporating bi-temporal data into their triple stores (even if this risks incompatibilities if the need arises to share data across parties for regulatory requirements); should semantic database vendors pick up the task? Says Allemang, “I think at the panel we’ll get people in the industry to challenge the business needs, and from the data side we can talk about technical possibilities, and we’ll see what we can bring together.”
You can register for NYC SemTech here.