I’m currently in the process of wrapping up my paper about Flickr the Commons. Retrieving a large dataset really has been beneficial in regards to my research. I’m not going to blog any spoilers but it is fun to look at the data especially in regard to the location of the Flickr Community members. To get a better understanding who contributes to the Commons we analyzed 167,871 accounts. This was the total amount of accounts who actively did something with the content of the institutions. One of the more interesting things was to see if we could determine the users locations and see which countries participate more.
Popescu and Grefenstette (2010) propose methods for analyzing textual metadata associated to Flickr photos that unveil users’ home location and gender. By exploiting tags with place named added by the user when he/she uploaded the photo, Popescu and Grefenstette were able to identify users locations. This method can be used when people take and tag photos of their home city frequently enough. Errors verification is difficult but errors probably occur for users who: do not photograph their home town; frequently travel to a same location; live between two cities; moved from one city to another without updating their profile.
The location field is an open text field where the user has no limitations regarding text entry. Users enter locations in various formats and languages ranging from exact geo-locations to descriptive locations such as “somewhere on Earth”. Because of this ‘creative’ use of the location field it was a difficult process to determine accurately determine the users location i.e. country and city. It was quite easy to determine that accounts with the location field US, USA, U.S.A. or US of A, belonged to the country United States, however, the process itself was quite time consuming and prone to subjectivity.
To overcome this problem we used Yahoo! PlaceFinder a geocoding Web service. PlaceFinder is able to recognize a large number of place formats and returns rich geographic data about each result, including geographic coordinates and address components (Yahoo!, 2012). From the 167,871 accounts, 61,607 accounts contained information in the location field. We used this information as input for Yahoo! PlaceFinder which returned the country, state, province and city it could associate with the queried location. In total 61,180 locations could be determined by Yahoo! PlaceFinder. This methodology worked quite well but had one notable drawback: in some cases Yahoo! Placefinder was unable to accurately determine the location when a user filled in information such as “Born and bred in North Berwick, East Lothian but now living in Edinburgh, Scotland”. In most cases Yahoo! Placefinder chose the last location and returned the respective information. Yahoo! Placefinder also had difficulty when it came to locations simply stating a city such as “Athens”. In some cases it would refer to “Athens, Greece” and in some cases “Athens, OH, United States”. While not 100% accurate this methodology gave us a rough indication of where the active users live.
It took some time but we could make an overview of the top 20 countries that tagged content from Flickr the Commons
|Country||Amount of tags||Amount of users|