This is interesting but I would use caution towards vague spatial notions, such as "Downtown".
There was a study by Dr. Dan Montello and colleagues on this matter: http://geog.ucsb.edu/~montello/pubs/downtown.pdf Basically what the behavioral experiment delivered is a kernel-density-ish region of where "downtown" is without defining clear boundary. On the other hand, if this opengeocoder.org could be a platform for building spatial models (similar to experiment collection in Montello et al, 2003) for vague spatial concepts, that'll be awesome. Best, Sen On Thu, Feb 16, 2012 at 4:21 PM, Laurence Penney <[email protected]> wrote: > Interesting idea. Scary number of man-hours till it’s comparable with > OSM/Nominatim or Wikipedia. Just a couple of quick notes: > > Some guidance about synonyms and homonyms is needed really early on. I > don’t want to add Reading, "Reading, England", "Reading, Berkshire, > England, UK" etc. and then worry about other Readings. Currently I cannot > even duplicate existing bboxes to achieve this stuff. Without this it feels > too much of a toy to spend more than 5 mins with. > > The worker bees deserve high-level public reassurance on the PD-ability of > the project from MS. If suits at some point decide/discover the imagery is > not PD after all, the project’s wrecked. > > Of course if people enjoy it there’ll be demand not just for arbitrary > polygons and lines, but also tags and a renderer :-) > > - L > > On 16 Feb 2012, at 23:23, Steve Coast wrote: > > > Hi > > > > I figured this is a good group to give a peek at something I worked on > last month: > > > > http://opengeocoder.net/ > > > > The premise is that a typical geocoder uses a large chunk of code to > import a large database in to a large geocodable database. Then another > large chunk of code is used to actually take a string and geocode it > against this large imported dataset. At the end of all of this all you’re > typically doing is showing some bbox for some string like “London” which > the user typed in. > > > > What if we did away with all that? > > > > Therefore opengeocoder is a simple list of strings and matching bounding > boxes. It has a trivial interface to let you add or fix existing geocodes. > It has an API on the side to provide results to 3rd party sites. If it is > unable to help you with a query then that string is saved and available for > anyone to later fix. > > > > The major differentiators against other sites are that the IP licensing > is clear, all bboxes are derived from imagery we have rights to, the bbox & > string data is put in the Public Domain. It’s trivial to use. The API saves > misses for later fixing. It’s hard to find a site that does 2 or 3 of those. > > > > Assuming this is interesting, there are multiple possible future > directions. Bootstrap with some PD data, allow points as well as boxes, > allow more complex polygons, a stronger API than just a JSON endpoint. > > > > Feedback welcome. > > > > Steve > > _______________________________________________ > > Geowanking mailing list > > [email protected] > > http://geowanking.org/mailman/listinfo/geowanking_geowanking.org > > > _______________________________________________ > Geowanking mailing list > [email protected] > http://geowanking.org/mailman/listinfo/geowanking_geowanking.org > -- Sen Xu Ph.D Candidate Department of Geography, GeoVISTA Center, 302 Walker Building Pennsylvania State University, University Park, PA 16802 http://senxu.net
_______________________________________________ Geowanking mailing list [email protected] http://geowanking.org/mailman/listinfo/geowanking_geowanking.org
