Wednesday, July 30, 2008

ChiefMall Contractor Search Engine

Alex Ksikes, who I worked with on GoogleModules.com some years back, this month with his team mates launched ChiefMall.com. Alex tells me “The search engine indexes a fairly large corpus of over 420 000 contractors and features geocoding and live indexing ... An interesting aspect of the site is that users can not only search for contractors but also post jobs to all contractors they have searched for. Behind the scene contractors are being matched based on their geographical location and expertise with respect to that job.” I asked Alex about more technical details behind the sites and he sent in this (edited to integrate links):

First of all I used the geonames database and built a simple geocoder around this. Because the zip codes of geonames were unprecise I also had to use the zipcodes.com zip-codes.com database. I used Sphinx for the retrieval engine. Live indexing was implemented using main + delta scheme where main are all the prepopulated data (state data) and delta the new companies being created. The state data (all active licensed contractors from California and 8 other states are featured) was grabbed from each board of contractor’s website (for example california is www.cslb.ca.gov). This information is all public domain and readily accessible. The data from each state was then parsed and merged into one master database. An interface was built to group all the state categories into one uniform category system. In order to return more results, synonyms were also associated to each category. (...)

Job matching is performed by extracting important keywords from the description of that job and performing a search restricted by the location specified. Once a job matches an email is sent to the contractor with the user information and job description.

The web site uses webpy as a framework. It’s simple and flexible and easy to hack into. The web site was designed in an MVC way. All the backend work was programmed in python.

The site also features an admin that was generated using a program that makes on the fly gmail looking applications. It creates all the templates and add the skeleton functions to be implemented. This program called auto-app-interface will be released open source once it’s cleaned up.

ChiefMall Contractor Sear ... by Philipp Lenssen | Comments (4)

>> More posts

Advertisement