You may recall my write up of Transparensee a while ago, I liked the approach founder Steve Lavine took to “fuzzy” search of structured data. Today I spoke with Steve again and he showed me a demo of his technology working on a real estate database. It’s very much worth checking out, it’s a cool visualization of how to make the web more useful. Head here, then fill in a few of the variables – price, zip, etc. Steve suggests 60014 for the zip, I think that database (greater Chicago) is populated. Click “enhanced search.” The sliders are the cool part, as are the similar properties. This kind of technology had all kinds of applications….
PS – this works on the Mac only in Safari…
The demo should also work using Firefox on the Mac, as well as IE 5.5+ and Firefox on Windows.
Maybe you meant 60614 as the Zip Code. 60014 isn’t a Chicago zip and doesn’t work.
Not sure anymore, to be honest! Steve, any insight?
60014 is just outside Chicago. This demo is sitting on a dataset of real estate from the Chicago suburbs. You can see exactly where 60014 is here: http://maps.google.com/maps?q=60014&spn=1.924204,4.122620&hl=en
It works just fine for me. It may be that you pressed “search” instead of “enhanced search” when trying the demo.
There are two buttons at the bottom of the search form. “Search” performs a standard SQL query. Don’t be surprised if you see “no results found” after hitting “search”. That’s a common problem when sending an SQL query to a database.
After trying a “search” and getting no results (or few results) click on “enhanced search” to see the difference that Transparensee makes. Now you’ll see exact matches (if any) followed by close matches. You’ll also see the sliders that John wrote about.
Here are some sample queries that should give you a sense for how our system works:
Query 1: Price = 250000; Zip = 60014; Bedrooms = 4.
Results:
“Search” – No results found.
“Enhanced Search” – Many results found that match perfectly for bedrooms and zip, but not for price.
Try making price more important (with the sliders) and zip and bedrooms less important. You’ll see that now you’re shown examples where there’s just one perfect match (instead of two) but the two imperfect matches are extremely close to what you want. This may be better where the end user is concerned.
Try this query:
Query 2: Price, Min=225000, Max=275000 (leave the first box blank; that’s the desired value; here we’re just using a range); Zip = 60014; Bedrooms = 4.
“Search” – One result found.
“Enhanced Search” – One perfect match found; many close matches found.
You’ll notice that the close matches are all in the price range specified (since that’s treated as boolean in this example; the min and max are hard stops) but that the system is trading off between close matches for zip and bedrooms. By moving the sliders above those columns you can change the search ordering.
Now try clicking on a property. On the right you’ll see a list of similar properties based on the metadata. If you select “show details” you’ll see why they’re similar (not all metadata used to calculate similarity will be shown, but some will).
You can even type a keyword into the text box above the “show details” link and you’ll see similar properties that contain the keyword you typed in (this demo doesn’t have a very rich free-text field, however).
Feel free to e-mail me with any questions you may have.
Steve