Homer tries to make sense of mapping
Dan Catt at geobloggers blogged over a week ago about Flickr about how they were using WOE ID to generate polygons of places that people have tagged in Flickr. A couple people have emailed me on how worthless this exercise is and how arbitrary it is. The Flickr Developer Blog goes into some more detail about the how and why Flickr is doing this and I think that should give everyone who doesn’t understand why this is important some better context. Personally I find it extremely interesting to see how people understand what place they are taking a photo at. A lot of this can all be tied back into Neogeography and what makes an expert. I’d wager people are more careful to geotag their photos in areas they are familiar with and less so when they are traveling. I don’t have enough photos tagged in Flickr, but I’d love to see maps from some larger Flickr users to see how they geotag the world.
What Flickr has is a way to visualize how people are geotagging their photos and they appear to be learning about to improve the process. I think this is a great application of technology to help better understand how humans perceive location.
10 Comments
“A couple people have emailed me on how worthless this exercise is and how arbitrary it is. “
It wouldn’t be neogeography without hordes of people bitching about their declining relevance.
On the contrary, it sounds borderline meaningless. A photo comes in with coordinates. Flickr uses those to “reverse geocode” some WOE identifiers onto the photo. Flickr then runs clustr to re-create shapes based on WOE identifiers and the coordinates. It seems like, over time, the shapes will simply converge on the original shapes used to do the “reverse geocode” in the first place (except that clustr doesn’t hand discontiguous areas or holes). The only flies in the ointment would be the occasional photos where someone actually hand-added a WOE.
Paul – I had the same concerns as you, that essentially the best shapes they could end up with would be approximations of the WOE source shapes.
However the project is much more interesting when you factor in the corrections process that Flickr is using – if they reverse geocode the location of your photo to some nonsensical place, you can fix it. These shapes are boring for big well-known places, but they’ll really come into their own for well-photographed but poorly mappped places and for places with several different names. Looking forward to it!
Paul…. On a micro scale this really is noisy, but on a macro scale I think it really brings to light what people are doing with location on Flickr.
This explains something I encountered on Google Maps yesterday. My wife asked where the La Brea Tarpits were. I knew there were somewhere in LA, but not exactly where, so I searched on Google maps. Result A was clearly wrong (wrong end of Wilshire), and someone had requested it be removed as a duplicate (though it did have pictures and 20 reviews attached to it). Result B was a set of flickr recorded locations from photos of the tar pits; unfortunately scattered across the city in 4 locations. Result C, pushed down under the 4 flickr locations, was the La Brea Tar Pits museum. It no reviews, no photos, and was listed as unverified, but it was the correct location. On the other hand, if I search for “Rancho la Brea Tar Pits”, I got only one result that was dead on. User generated content is also dead on for all 33 results.
We at Abaq.us have linked our service (www.MyGeoDiary.com) to Flickr such that users can easily use their gps tracks to geotag their photos and upload the resultant geotagged photos to Flickr. We try to make the process easy by trying to do a pattern match on timestamp. We are going to add reverse geocoding (place-cloud) feature soon. WOEID is a cool feature especially in its fine-grained control on privacy settings. But we couldn’t get Yahoo’s geoplanet Platform (for WOEID) to work with lat/longs. So will use Google to reverse back into a place info with the recorded gps track data (for now).
Full Disclosure: I work at Abaqus
I think the interesting part is for WOE elements that don’t have shapes, like unincorporated populated places. Correct me if I’m wrong, but I believe that Yahoo just uses the center point and a rough radius for these items. I would like to see what flickr can make for these places – since information on their actual boundaries is very scarce.
Or adjust the centroid of each WOE ID for better accuracy. As I said this could really give Yahoo! some great feedback.
Does anyone know if someone has put up a mashup where you can view the shapefiles they’ve got on a map? or is the only way right now through the API?
Flickr blog just posted a guys app that does what I was looking for. Just wanted to post here to follow up on my own question
http://boundaries.tomtaylor.co.uk/