Dopplr gets a gazetteer upgrade
This week, Dopplr got a big upgrade in the way it handles places. Firstly, we’ve expanded our coverage of locations to nearly 200,000 cities worldwide. We now list every place that has a population figure in Geonames. We’ve still got a way to go in coverage, but it’s a huge improvement. We’ve also reworked our city URL structure and our search pages. People of Llanfairpwllgwyngyll, rejoice.
The bigger the location database gets, the more potential there is for complexity in our UI. We’ve handled this with some rather pleasant and appropriate uses of Ajax and geocoding services. When you add a trip to a place like ‘Oxford’, ‘Cambridge’ or ‘Springfield’, we check your trip history and the most popular trip destinations and make a suggestion – perhaps you mean “Cambridge, United Kingdom”?. If we’re really stumped, we turn to Google’s geocoder, like so:
- Google, where is “Sherman Oaks”? (Sherman Oaks is a district of Los Angeles)
- Google says “Sherman Oaks, CA, USA at 34.15116,-118.44456″
- We look in our database and ask, what’s the nearest thing to that location that we know?
- We find Beverly Hills, CA, USA at 34.0736,-118.4 and offer it as a suggestion.
This also means that you can add trips using any location style that Google knows about – postcode, zipcode, airport IATA code and many more.
UPDATE: We’ve made a few more refinements in response to comments below.
Congratulations on finally adding Brisbane, Australia!
But where is Hobart, Australia? It’s also a state capital.
And of course, we’re still waiting for Hong Kong (rather than “Hong Kong (historical)”.
You still seem to be missing the fairly substantial city of Winchester here in the UK. I’m thinking that you have a bug somewhere, because I’ve noticed more than one friend/colleague HAS been able to set this to their home city.
A search for my hometown, “Hualien” on Geonames gives me this result:
??
Hua Lien Kang,Hua-lien,Hua-lien-chiang,Hua-lien-hsien,Hua-lien-shih,Hualian
Which is listed as a “populated place”, but I don’t see it yet in Dopplr, using any of those names.
A search for my hometown, “Hualien” on Geonames gives me this result:
[Chinese characters deleted because your blog doesn't support unicode.]
Hua Lien Kang,Hua-lien,Hua-lien-chiang,Hua-lien-hsien,Hua-lien-shih,Hualian
Which is listed as a “populated place”, but I don’t see it yet in Dopplr, using any of those names.
I’m sorry, I should have been clearer. When I said “has a population figure”, I mean that they list an actual number for population – for example, in a search for London they say “population 7,421,209″ in the search results. They don’t have this figure for all 2 million or more populated places – only about 200,000 in fact.
We’re now working towards using the entire Geonames database, but we’re not there yet. The difficulties include how to rank search results (currently we use population figures as part of the heuristics for making suggestions) and how to build a UI that shows so much data without turning everything into long scrolling lists.
We’ll get there before long, I promise.
[...] that understands the Webcal scheme). Best of all, it now knows a bunch more places – nearly 200,000 cities, [...]
Targ, Andrew and Kerim,
You can add the correct number of inhabitants on the http://www.geonames.org website to make sure your cities will be supported by dopplr with the next data load. Not only will this help dopplr, it will help all applications using GeoNames data. There is a simple wiki-style edit interface on top of google maps.
Matt,
You could also load PPLC (capital of country) and PPLA (seat of province) even if we don’t know the exact population since those are usually important cities (Hobart and Winchester from the postings above are PPLA).
Regards
Marc
Marc, that’s an excellent idea about the capitals. I should have thought of that before. Clearly it’s ridiculous for Dopplr not to list a capital city. I’ll get to work on it.
This is really terrific. As someone who lives someplace that is still too small to be on Dopplr, I appreciate being able to move my home town a few miles closer to where I actually live and have edited the population details for my town so that maybe you guys will grab it on your next data load. For anyone wanting to know how to do this, it’s as simple as going to geonmaes, logging in, searcing for your location and clicking on it on the map [I didn't see another edit link but I may have missed it] and then you’ll see an edit link where you can add population details.
[...] el blog de Dopplr, me puedes encontrar con una de esas situaciones inesperadas que pueden surgir cuando diseñas un [...]
I just added a trip to Tucson, AZ, USA. But Dopplr only “knows” about South Tucson despite Tucson being the number one location when one searches on that name on GeoNames.org.
Augment to my post: Dopplr defaults to South Tucson, but offers a link where 4 different possible selections are shown including Tucson. But if I select Tucson, then Done, I get a Dopplr 404 not found screen indicating “That page is in the interzone.”.
“You can add the correct number of inhabitants on the http://www.geonames.org website to make sure your cities will be supported by dopplr with the next data load.”
I did this back in August for a number of locations, eagerly anticipating that dopplr would then allow me to plan future trips to more accurate locations (and to get a decent home location). However, dopplr doesn’t seem to have reloaded the geonames data, is this going to happen soon, or am I wasting my time?
Jim,
Our update process for geonames is still a manual process. We’re working to automate it, but for now it only happens every few months. We’re planning another update before the end of the year.
Matt.
[...] They use place names drawn from Geonames, and seem to have most major cities covered. (See posts here and here on how that autocomplete has [...]