Results should be displayed in Geo / Location order first

[GoogleCodeExporter]

The system should allow results to be sorted by geographical location.  In 
fact, this should be the default order of the results on the All For Good 
site.  I understand that ‘location’ searches are converted to lat / long 
with a default of 50 mile radius.  The current system uses the keyword 
match as the primary ordering.  This should change to the location 
parameter determining result ordering.  For example, if I enter a zip Code 
as my location preference the results should be ordered in sequence by 1. 
Actual Zipcode matches, 2. City, 3. State, 4. National, 5. all other until 
the number of valid results is returned.  The same should happen when 
searching on City or State.

            Volunteering is largely a geographical process.  Except for 
natural disasters, most volunteers search for opportunities within their 
area.  Driving 1000 miles to work on a weekend project is not practical.

Original issue reported on code.google.com by stewarts...@gmail.com on 23 Dec 2009 at 12:25

[GoogleCodeExporter]

hi!

in general, most parameteric (more than keyword) search engines offer ways for 
users 
to change the sorting order, and 'relevance' is just the default.  Keep in mind 
that 
we use ranking to make up for our lack of quality control-- if you sort on 
something 
other than 'relevance' (which blends geo-distance and other factors), then 
quality 
goes down fast.

In english:
1. we already incorporate distance as a "strong signal" in the ranking function.
2. if you sort *only* on distance, the results may not be usable.

However, I agree with the premise: users should be able to sort on distance 
and/or 
time.

Engineers: if you want to experiment, just pass in a "sort order" arg into the 
existing ranking code (scoring.py), then skip most of the code if sort != 
relevance.  
For release, we should test this with human eval by creating a server instance 
with 
sort-order hardcoded to the new value (one instance per sort order) then 
pointing 
people (mturk) at that instance, and comparing the scoring results-- IMHO we 
shouldn't launch if perceived quality goes down the toilet, esp. after the 
major 
improvements from the last round.

Original comment by adam.sah on 23 Dec 2009 at 4:39

• Changed state: Accepted

[GoogleCodeExporter]

oops-- this google code project is old: moving to allforgood.

Original comment by adam.sah on 23 Dec 2009 at 4:41

• Changed state: Invalid