The long-promised “semantic” web is starting to take shape

Improving the internet

The web: some antics

Aug 28th 2007
From Economist.com

The long-promised “semantic” web is starting to take shape

FOR all the tricks that the world wide web can perform, it still resembles a collection of one-trick ponies rather than a concerted cavalry charge. You can book an airline ticket, hire a car at your destination, arrange concert tickets for the evening that you arrive and even get directions from the airport to the concert hall. But you have to do it all yourself, one element at a time. You cannot delegate the process to a website as you might delegate it to your secretary or your long-suffering spouse.

You can, however, delegate some things. At least, you can if you are Rael Dornfest, a technologist and entrepreneur from Portland, Oregon. When Mr Dornfest e-mails his business partners about meetings and interesting titbits worth archiving, he copies the e-mail to his assistant, Sandy. Though she cannot yet organise his evenings in foreign cities, she can run his diary. She also runs his address book and forwards reminders from his wife to his mobile phone without being asked.

Sandy, of course, is not a person. She is a piece of software that Mr Dornfest and his colleagues are developing and whom, once she is thoroughly tested, he hopes to sell to a wider world. She is one example of a long-promised technological advance: the semantic web.

The semantic web is so called because it aspires to make the web readable by machines as well as humans, by adding special tags, technically known as metadata, to its pages. Whereas the web today provides links between documents which humans read and extract meaning from, the semantic web aims to provide computers with the means to extract useful information from data accessible on the internet, be it on web pages, in calendars or inside spreadsheets.

It does so using a trio of new technologies: the Resource Description Framework (RDF), the Web Ontology Language (OWL), and the SPARQL query language. Together, they allow computers to group objects and their features—from prices and measurements to locations and user ratings—into meaningful relationships and hierarchies, by analysing their associated metadata.

The idea is that eventually such metadata will be incorporated into every web page and electronic document. But that is not the case at the moment, so a further layer of software is needed to infer the metadata from web pages, e-mails and other electronic documents. Whatever their origin, the metadata labels can then be used to do useful things. A piece of software can, for example, compare goods that are similar but not identical and then recommend the best (or the cheapest, or the best value for money) to a potential customer. In the field of travel, attaching metadata to everything makes it possible to link up airline schedules, car rental and hotel bookings.

If this sounds eerily similar to the kind of thing a human travel agent used to do in the days before the non-semantic web almost killed his profession, that is exactly what Gregg Brockway has in mind. Mr Brockway is the co-founder of TripIt, a firm based in San Francisco. His intention is that people should be able to dump all of their travel details (electronic tickets, car-hire bookings, hotel reservations and so on) straight from any reservation site into a central repository, which TripIt will run. Then RDF, OWL and SPARQL—or, at least, TripIt’s implementations of them—will sort the information. The software will group the data appropriately and annotate the result with weather forecasts, driving directions, restaurant recommendations and even the travel plans of friends and family. It will then send the results back to the user—or, in the case of the driving directions, directly to the car’s navigation device.

Another area where the semantic web may make a contribution is personal finance. Even if they have not heard the term, most people will be familiar with the idea of what a company called Wesabe refers to as “bank puke”. This firm, which is also based in San Francisco, plans to make money by clearing up such puke and turning it into useful information. The idea is that its customers will be able to feed their bank statements, credit-card accounts and so on into the system as if they were throwing reams of paper onto an accountant’s desk.

Wesabe’s software sifts through all the transactions and makes comparisons between users. It can then do some of the things that a human financial adviser might, such as recommending to a customer a different car-repair shop if other customers in the same area are using a cheaper one. Not yet, perhaps, a cavalry charge. But not a bad performance of formation riding.