[Developers] Google Custom Search Engine + Vancouver Base

[Jim Pick]

It shouldn't be too hard to parameterize a few parts, and make it 
generic.  It doesn't use any custom types from the Vancouver base.

One thing I noticed is that I need to sanitize the data a bit better 
before passing it on -- any poorly formatted URIs that get added to the 
base will kill it for everybody.  I bet there's probably already a 
JavaScript library out there somewhere to do that.

The custom search engine needs to be created and registered first with 
Google (tied to a user or group of users).  They do have an automated 
way to generate custom search engine descriptions.  Also, they require 
advertising, unless it's for a non-profit or charity.  They do pay 
AdSense revenue to whoever has registered the search engine.

So in order to make it generic, some decisions have to be made about how 
to register and configure it on the Google side (including decisions 
about who gets the AdSense money, however small that may be).

Google also has limits on the number of "annotations" (aka. sites) they 
will index.  I think it's 3MB per annotation file, 10MB max.  That 
limits the total size.  I'm at 132K for 1168 URIs now.   If I'm 
aggressive in collecting data, I might eventually hit the limit for the 
Vancouver base - at which point I'll have to get selective.

Yahoo BOSS looks like an interesting alternative to try out as well 
(with news and images too).  There probably are other similar services 
lurking out there (eg. sites that read OPML files).

Cheers,

  - Jim

Robert Cook wrote:
> This is very cool!  Jim, if you do the small tweak so it can work with  
> any base, we'll weave it in to the interface.
> 
> R
> On Dec 8, 2008, at 5:42 PM, Jason Douglas wrote:
> 
>> Very clever!!
>>
>> I never thought of that mash-up before, but the results speak for
>> themselves.  :-)
>>
>> -jason
>>
>>
>> On Dec 8, 2008, at 5:33 PM, Jim Pick wrote:
>>
>>> Hi,
>>>
>>> I just used Acre to build a Google Custom Search Engine based on the
>>> data contained in the Vancouver Base.
>>>
>>>  http://vanbase.jpick.user.dev.freebaseapps.com/search
>>>
>>> Please look the Acre source code and feel free to clone it.
>>>
>>> There are currently 1145 URLs it searches against.  I've only done a
>>> few
>>> searches with it, but it does a very good job for things in the
>>> Vancouver area.
>>>
>>> Cheers,
>>>
>>> - Jim
>>> _______________________________________________
>>> Developers mailing list
>>> Developers at freebase.com
>>> http://lists.freebase.com/mailman/listinfo/developers
>> _______________________________________________
>> Developers mailing list
>> Developers at freebase.com
>> http://lists.freebase.com/mailman/listinfo/developers
> 
> _______________________________________________
> Developers mailing list
> Developers at freebase.com
> http://lists.freebase.com/mailman/listinfo/developers