Quote:
Originally posted by Nicky
we can start using the city name we live in much more on out porn sites......"this bitch is from denver, where she sucks cock daily"
|
Maybe you were kidding, but I seriously doubt that in itself would work. Google's almost certainly using actual physical address information (from a InfoUSA or the like) and tying it back to website information as best they can. Unless you're gonna say "this bitch is available for your escort needs at 555 High Street, Denver CO 30103", Local Google probably isn't gonna touch you.
Dun & Bradstreet (among others) has been trying to do something like this at the data level for a long time now, but they never had a good enough data store of addresses parsed out from websites to make it more than 10-20% effective (in terms of coverage of their US B2B databases). They've got a great index of businesses, but their ability to reconcile that to actual websites has been spotty at best (at least it was as of a year ago when I inquired into leasing their data).
Just a few more self-indulgent random blatherings (local search is near and dear to my heart... ) :Graucho
Search matching:
LocalGoogle has most likely taken a tach different from D&B or others who've dabbled on this front before. They're clearly way ahead of the game on mining address information (and of course keywords) out of websites and other data sources. They've got the web page index, and they're using it (again). If you've got "pizza" or "ethiopian restaurant" and a usable address on your page, they're most likely serving your results in the appropriate way.
I like that they're doing some very basic (non-intrusive) spell checking and correction. I'm seeing that you can misspell lots of basic terms (try 'piza' instead of 'pizza') and you'll get first the sites containing the misspelled 'piza' (a reward to the spelling-challenged among us :)), then all sites containing the corrected 'pizza' below it. Smooth, non-threatening, but helpful.
Geolocation: Looks like their process is good enough (not hard to do - lots of algorithms readily available these days - at least for the US -- check out opensourcegis.org if you give a shit) -- seems that they're doing it based only on ZIP codes, not on radius from the center of your search (inefficient, but tolerable). Could be a planned enhancement tho -- they've got a CGI arg in their results URL where radius=0 for every results set I was able to generate. Biggest disadvantage to this approach is that ZIP codes in the US (and most other places) are inherently inefficient -- they're misshaped and bound more by political gerrymandering or outdated population estimates, instead of by actual distance from a point or, better still, traffic data assessments for travel time from point A to B.
Purpose of Local Search:
I took a bit of liberty with that last statement -- I assumed that the purpose of local search is to find somewhere that you want to *go*. That's certainly not always the case -- oftentimes you may just be looking for a phone #, etc. That's the easy stuff for a search engine -- if the data is available. To me, wWhere the smart stuff comes into play is in the ability to tell me how to get to the local search result location I'm interested in. But that gets into integrating GIS technology with local search, and that's not what they're trying to do (yet).
When you do a city-based search (e.g.;l 'restaurants' in 'los angeles, ca'), they use a mappoint (latitude/longitude) which appears to be at an arbitrary point within the city - city hall or the like. Something they can use as a fixed mappoint. Nothing very novel about this -- I seem to remember that the guy who won their 2002 programming contest basically wrote this for them.
I think what bothers me most is what they didn't do with this project (yeah, I know it's beta, but this is kinda block-and-tackle stuff). Local search, at its core, wants to grow beyond the Yellow Pages (or whatever color the B2C telephone directory is in your part of the world). Google's stopped short of that -- the only results you'll get are businesses with some sort of web presence containing the search term you, as a consumer, would use to find them. Reality: Mom & Pop's Liquor Store down the street isn't going to have anything more than a billboard webpage (at best) anytime soon. Reality #2: I want to find them, but if I can't remember their name (or if Google/someone doesn't have them indexed with better metadata), I'm still screwed.
Unless someone gets smarter about how local business web pages get built -- sprucing them up with more keyword-enriched details and information about the business itself or, better still, categorical, parametric, and *standardized* metadata along the lines of a consumer-focused SIC code system -- we're still better off using a yellow page search tool. At least that's going to be a lot more complete than a purely algorithmic product like this.
The real hippykiller here will be when someone does what I'm yapping about above, but also ties it into those big brother traffic cameras posted at stoplights and on freeways to tell me the most effective way to get to where I want to go *now*. Feed the information to the Wide-Fi-connected GPS nav system in my car (powered by my car's hydrogen-powered engine), and have the system effectively recommend the best route for me given time of day, historical traffic patterns, current traffic conditions, etc. Maybe then local search (and big brother) will both be a little more useful.
</ rant>
OK, back to slingin' porn now... :Graucho
<br>
