The Yahoo! Geo Technologies Blog

Welcome to the Yahoo! Geo Technologies blog, your place for Place.

At Yahoo! we’ve been using geo technologies to ensure that our half-a-billion users globally receive the most geographically relevant information possible. As part of the Y! OS initiative, we are making these geo developer tools accessible to all, and writing about it here.

Here you will learn about Yahoo!’s products, tools, and resources that help bridge the disconnect between the Real World and the Internet, including coverage of Yahoo! Maps, our geolocation and geoinformatic initiatives, user location services, and all clever uses of geo technology at Yahoo!

We’re therefore taking the opportunity to proudly announce that today we are launching Yahoo! GeoPlanet™, an open, permanent and intelligent infrastructure for geo-referencing data on the Internet. (You may have seen this previewed as the Internet Location Platform at Where 2.0 in Burlingame, CA earlier this year.)

Our two driving principles in creating GeoPlanet are to be as comprehensive as possible (we continue to add thousands of places daily) and to ensure that we capture the geography of the Earth as it is called by the world’s people. Within the collection of over six million named places we include (big breath): Continents, Countries, Counties, States, Provinces, Prefectures, Country, Regions, Federal Districts, MSAs, Provinces, Parishes, Departments, Districts, Communes, Municipalities, Districts, Wards, Cities, Towns, Villages, Hamlets, Postcodes, ‘Supernames’ (USSR, Western Europe, Latin America), Time Zones, Points of Interest, and Colloquial Names (such as Wine Country, the French Riviera, South East England, SOMA, and the Pacific States). Phew.

What’s more, we’ve attempted to solve one of the real bugbears of geographic indexing: how do you uniquely identify places in a uniform and consistent manner? We’re talking geotagging here: string matching won’t always help (there are over 100 Springfields) and providing a string geographic context will not always work (there is more than one Wayne, PA, USA for example). Lat/Long is of course the obvious choice, and this is perfect for geotagging (say) photos that were taken at a single point on the earth’s surface. The coordinate-based approach however can fall down when we want to associate a unit of information with an area – such as a country, region, or neighborhood – because we usually do not know the exact point within that area that the contents of, for example, a newspaper article refers to.

This question therefore is how do we associate the spatial and political entity with a news article or other information about that place? We can geotag a newspaper article about Afghanistan with its centroid, but this associates the article with a point on the earth’s surface – we don’t know which intersecting place it specifically relates to: Afghanistan, Kandahar, or a specific neighborhood of Kandahar itself. We can of course represent the place by a polygon, but this becomes hugely cumbersome. More critically, it becomes difficult to match with articles geotagged in the same manner. For different systems to geotag two different articles about Afghanistan, they would need to employ identical coordinate pairs to represent that place – just one pair amiss and the systems are geotagging two different areas. We know that this can be determined though various spatial functions here — our point is that we should not need to; fundamentally, coordinate pairs are best employed to describe Space, not Place.

What we still require in this scenario is the ability to geotag an information unit with a unique identifier for a place, so that the information is associated with a Place – or in a quantum-like manner, associated with all points in that place at the same time. Yahoo! provides what we believe is the solution with WOEIDs, Where On Earth IDs, unique identifiers for every named place in GeoPlanet. When you tag a unit of information with a WOEID, it associates that information with the concept of that place, not with a spatial approximation of the place itself. This is actually much more appropriate — our concern is to provide a common naming convention, and to ensure that places are correctly represented in relation to each other in a global, consistent framework. In practice this means that we are not in a position to claim that a particular neighborhood stops at one block and starts at the next, only that the concept of that neighborhood be identified consistently. Our primary concerns are relative geography and the semantics of place.

Lastly, because we conceive of the idea of a place as being conceptually distinct from how it is called, we can ensure that multiple names for the same place are managed consistently. For example, München in Germany is Munich to the English speaking world and Monaco di Bavaria to the Italians. But it may also be keyed as Muenchen and Munchen if special characters, diacritic marks, and ligatures are not available to the user. All of these spatial appellations are simply multiple names for the same place, and therefore reside within GeoPlanet mapped to the same WOEID (676757).

We are delighted to present GeoPlanet to the Geographic Developer Community and look forward to posting further news and musing on Yahoo! Geo Technologies here shortly. There’s much to talk about.

The Yahoo! Geo Technologies Group

This entry was posted on Thursday, June 5th, 2008 at 2:09 pm and is filed under Site. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.