@ Garage Sale











About @GarageSale.com

 

 

AtGarageSale.com is a vertical search engine focused on collecting as many yard sales from the web (users can also submit sales directly here), and providing a simple interface for finding the sales near you and sales that contain items you are looking for.

 

 

This sounds simple enough.
The problem is that most yard sales on the web are presented in unstructured form, meaning that locations and dates of sales are not explicitely given, but are incorporated into text in numerous ways. Without date and location information separated, it is hard to filter out the sales on particular date and those that are near you.

 

 

So how is this problem solved? First, a part of text which resembles address needs to be detected, and then that part analyzed and verified whether it is really address. Similar procedure is done for dates. We are working on detecting times as well. Anyhow, as an illustration of the diversity of address formats, here is a list of few successfully detected addresses:

 

 

951 OSBORNE ST CA
3585 N Lee Trevino TX
Carmel Mountain Ranch, 92128
Indian Creek -- Hampden Twp PA
8301 55th Way Pinellas Park FL
3130 Boxelder St, Deltona, Fl 32725
2826 O'Neill Drive, Bethel Park, PA
1507 S ANDREWS AVENUE IN FORT LAUDERDALE. FL
450 East Surfwind Drive Garden City 29576 SC
26111 W. Timber Ridge Dr, Channahon/Minooka IL

 

 

Some may ask what's the big deal, as above addresses may appear more or less standardized?! Consider few false or imperfect detections below, which you will not be seeing often in the future as the improved algorithm's accuracy and precision are around 90%.

 

 

1 low price FL
28 ESTATE SALE IL
3 Homes gathering FL
2 sets of Hardware FL
BRICK NJ DON'T MISS THIS ONE! PA
41 ave and north of Thunderbird  AZ
6100 block of North Austin Ave. on IL
7016 Ohio River Blvd., , Huge Variety PA
75 additional Home Furnishings and Utensils, IL

 

 

In coming weeks, not only will locations and dates be more accurately detected, but there will be even more sales in the database, as the work on a robust content detector is nearly done, and in symbiosis with a focused crawler many big and small yard sale sites will be indexed soon.

 

 

© atgaragesale.com