Companies considering how they might help speed the development of the semantic web by creating something useful for their audience can benefit from the experience of people who are doing just that -- people like Taylor Cowan, emerging solutions principal at Sabre Holdings, and Jay Fichialos, director of ideation and experience at Sabre Travel Studios, which identifies, tests and launches new travel concepts.
The two recently shared their experiences working on TripBlox, billed as an initiative to use open standards with a travel twist to help publish, promote, and aggregate trip ideas. The idea is that users can publish trip information on their own blogs, using XHTML or hAtom microformats and some semantic web markups, so that TripBlox can become aware of the existence of the formatted data. It will consume that information and pull it together with related information to become something of a search engine for trip ideas.
Advice from the trenches includes:
Aim for richness
Microformats such as hAtom, a microformat for content that can be syndicated primarily on web postings, are useful, but difficult to parse. If you're going to be exposing data on the semantic web, RDF "is the secret sauce, a rich, flexible way to create relations between things," Cowan said at the recent Linked Data Planet conference. People create information using the web tools they like the most, then TripBlox takes RSS feeds of that information from their accounts, such as Flickr, with some semantic markup in its links, consumes that, and stores it in an RDF triple store on the back end, to link things to do with people who are doing them with the location they are done in. "These are the building blocks for creating something that will be more interesting in the future."
Keep it simple
Fichialos said that was a lesson learned during the earlier development of a social networking site, a primary project the team was working on, in which so many of the assumptions made about what kinds of functionality people would need were wrong. The team spent months working on how they would match questions asked by the community using the site with the community of people who could answer them, and what they would do if no one answered them, only to find upon launch that there was an average of six answers to every question asked.
"So we put all that effort into trying to match and we found out that most of the time community members went to the effort themselves to look through questions to see what they could answer. The lesson was we didn't need to put as much effort into it and we could have launched earlier and gotten feedback. So we took that [experience] back to TripBlox and said to make it as simple as possible and get it out to users ASAP, because they will tell you where it breaks,"
Open beats closed
Fundamentally, TripBlox is about sharing. When there is a choice between dealing with privacy concerns or not sharing, the TripBloxy opted to share knowing they would potentially lose some of the audience. That choice also simplified things.
"When you deal with segmenting a lot of stuff out, that makes the whole system much more complex," Fichialos said. "So we defaulted to open data, open feeds that are available. We're still holding true to that," although as the product matures it is likely people will partition some of their own space. User management, in fact, is delegated to OpenID, a decentralized identity system that lets users use a single digital identify across the Internet.
Cowan said the key thing is to let the data flow and add value as it goes by, rather than trying to be a content manager and a content store. "In the past sites owned content, you can't get it out of that site, it was a one-to-one relationship," he said. "But we are trying to add a little value on top of content that is already on the way."
IE 7
Firefox 2.0
