The Semantic Technology & Business Conference has been underway since Sunday, with tutorials and lightning sessions catching audience interest. The conference presentations get underway today, most of them following on the heels of the opening keynotes given by Bart van Leeuwen, firefighter and architect at netage.nl; Jay Myers, web architect at Best Buy; and Steve Harris, CTO of Garlik, a part of Experian.
Best Buy, as readers of this blog know, has been diving deep into the semantic web waters under Myers’ direction for a few years now, and he shared that journey with the audience at SemTech.
In a discussion before the keynote, Myers said he was glad to have the chance to talk about how the Semantic Web and Best Buy's start of tinkering with Linked Data are bright spots for the retailer. It's been something of a rough year for the company, which saw the resignation of its CEO Brian Dunn and declining comp store sales, and now is moving to shrink its physical footprint and reduce its cost structure.
As it happens, the shift from physical stores to an even stronger online focus aligns well with the work Myers and his team have done to use RDFa and microdata with the schema.org vocabulary to grow online visibility. “That is so much more important to our business now, and people are more in tune with it,” says Myers. “With more business going through our online channel vs. going through our stores, being discovered and visible on the web is even more important.”
Myers has proven results with RDFa going back to 2009, when the work his team did led to a 30 percent increase in the amount of page traffic from sources such as search engines. But impressing people with ROI tied into SEO from using schema.org, a format that also adds the value of the Google brand, helps a lot in selling structured data markup to the business. “A layperson can get that,” Myers says. Even the front-end development staff he’s handed projects off to seem “a little more onboard with microdata.”
Myers has been at Best Buy nearly seven years, and it’s now that the structured data markup work of the last four years is really hitting the spots he’d hoped it would. ”Maybe you don’t have to do these elaborate user experiences but just publish data in ways that make sense to machines,” to help users find what they need, and consequently deliver to the business “pretty tremendous benefits from being open with data and marking it up with semantic web standards,” he says.
With the data in hand to show that “this stuff works,” the opportunity is there to do more proofs of concept around larger data, to explore ideas such as what insights can be extracted from triple stores and how Linked Data can be used to help change the customer-facing web presence.
As an example, Myers points to smart web services – imagine a service-oriented architecture for Best Buy that lets it pull a lot of different linked data sources into the web site, such that it becomes almost hyper-personal. For instance, for a certain zip code it can query a data source for the weather there, and then, with that data in mind, run a SPARQL query against the retailer’s product database to see what the most relevant products are for those conditions, and display them to the user.
“The search engines get insight from schema.org and RDFa. But now I can take that, do markup in a machine-readable way and do queries against that, and that’s where we get insight, not the search engines,” Myers says.
Another Semantic Web Journey
Similarly, van Leeuwen’s keynote took the audience through a journey, too, this one about using Linked Data to help in fighting fires in Amsterdam and Amstelland, and how lessons learned there can apply to their own semantic web efforts. The highlights to hit, according to an earlier discussion van Leeuwen had with The Semantic Web Blog, are less about the nitty-gritty technical details and more about making sure that everyone in an organization starts to see the benefits of linking data (as with Best Buy’s work, it’s about proving the concept).
“If you have people who frantically sit on their data silos, throwing the semantic web at them isn’t solving the problem,” van Leeuwen says. “They have to change their mindset as well.”
It’s possible to help them do that by starting small and showing results. “You can keep a lot of the systems you use on a day-to-day basis and just use back-end semantic web technology to make simple connections just between two systems, and see what happens when you tie them together with unique identifiers, and then maybe move to three systems,” he says.
In his own experience as a firefighter, keeping data silos spells trouble. It takes a fire truck about four minutes to get to an incident, and when data is silo’d, those four minutes are not spent as well as they could be in terms of having all the information needed to quickly assess a dangerous situation in real time. Knowing later that information was available that could have kept a colleague from injury doesn’t help anyone, after all. But projects that start small, as his did, can grow with more links among internal data sources that make important information accessible at the right point in time, and then link to open data sets as well, to make the graph of information bigger and bigger. And, quickly accessible to firefighters when they need it.
There’s more hopefully in store for the real-time emergency response application he's created for the firefighting force. He says a goal is to create a structure to really mobilize this information, so that data is still available when a network connection isn’t. There’s also potential value in hooking to social media data. For example, hashtags can be used for big incidents – a factory fire, say, where chemicals may be burning – and people can post images of plumes of smoke or other information to Twitter.
“Right now we have to send out teams to look for various angles of smoke and do measurements,” he says, whereas this could provide real-time data and even different angles to show alternate perspectives. “You can also create a linked data perspective on the incident you are fighting, with a tweet with a URI that shows factual information of the incident, but also links to some document describing how civilians should act – for instance, close the windows when there is black smoke.
You can actually use social media to inform people in the area of a disaster how to respond [in an automated way].”
The keynotes make one thing very clear: There is power in the Semantic Web and Linked Data. And there's opportunity for many others to gain that power, too.