timothy falconer's semantic weblog
Big Fractal Tangle


RDF
 



properties that pay

In 2000, my wife and I wanted a new house. We drove all over, we talked with our realtor, we went to open houses, we looked online. Nothing felt right. We decided to write up our "top ten" lists of what we wanted most. My list led with "space" and "light" and "land". Hers included "jacuzzi" and "closets" and "great kitchen". We both wanted "fireplace." These lists became our informal search profile.

Such profiles are the stock in trade for real estate agents, whether they're helping you buy or sell a house. The better a realtor can predict whether a prospect will like a particular property, the more money they'll make. Good realtors can gauge the attributes that'll close the deal. They encourage buyers to focus and finesse their search profile so they can more quickly match desire to opportunity.

As helpful as MLS (multiple listing service) database systems can be, deals often get made with idiosyncratic attributes that aren't in the database. Individual properties and realtors and regions have unique selling points. Lanai's are essential in Florida, but unheard of up north. Realtors want the flexibility to add new criteria to highlight competitive attributes. This is the first reason why semweb technologies would be great for real estate: flexible vocabularies mean realtors can pinpoint the newest "attributes of opportunity." They can say, "Hey, add 'massage room' to the list, it's becoming a must have in west Florida."

This brings us to the next good reason why semantic web technologies and real estate are a match: there's next to no consistency to the many MLS databases in use. MLS systems across America have varied fields with complex constraints, making the domain ripe for an ontology or two to help tie things together. Effective data integration will enable individual realtors to extend their reach, matching buyers to properties in other regions, using other systems.

My last good reason for real estate as a semweb domain is the potential for semi-autonomous software agents. I'd love it if our profile could travel the country in a decentralized, many-to-many manner. Even better, I'd love it if my cell phone or PDA would start to vibrate when I'm driving near a property that matches my profile. I could pull up a web services menu and choose from a menu that gives me details, directions, and contact info for the listing agent.

While this last idea may take a while, I think the rest of it isn't that far off. I believe real money could be made next year in this domain if someone took up the challenge.

Incidentally, my wife and I did find our dream house. It's got plenty of space, lots of light, and acres of land. It's got a jacuzzi, and big closets, and a great kitchen. It's even got a fireplace. Score one for our realtor, who found the place, our most successful "search result" to date.




Trackbacks (use http://immuexa.com/cgi-bin/mtype/mt-tb.cgi/72)


Comments

This is good stuff Tim. As the property industry have been early adopters of the web, I am surprised that they are not leading the way on delivery of machine readable data in either XML or RDF. How much better for the service for house hunters would be if new property notifications were delivered by RSS rather than Email. RSS seems a logical point of entry for this and many other industries. Once the first steps are taken in this direction, I am sure that the use of RDF to allow flexibility and extensibility in data would naturally follow. And once there is a decent critical mass of data 'out there' then interesting applications such as the throbbing phone based on a search profiles, semantic data servers and mobile technologies would follow as night follows day.

posted by Victor at December 22, 2003 11:46 AM


For this to work well, you need an ontology/vocabulary to dominate an industry. If every realtor comes up with his own, we are right back where we started. In a way, I hope creating one remains geeky to discourage amateurs.

There need to be tools to recode data to a new ontology when the selected one goes out of favour.

Definitions in the oncologies need to include type information, e.g. that "bathrooms" is in integer, not a descriptor like "indoor" or "large" or "ensuite".

Ideally definitions should also contain some guarantees on validity. E.g. that zip code contains only numbers issued by the US postal service, and that they are at least guaranteed to be in the correct state.

A gender field should let you know what possibilities to expect besides M or F.


posted by Roedy Green at December 30, 2003 12:08 PM