Event Detection and Visualization of VGI Khatereh Polous Technical University Munich (TUM) Cartography Department 11.06.2012
Background web 2.0 participatory platform colloquial readand-write functionality Individual users collaboratively generate the content
Background Text messaging, Social networks, Photos, Video, Blog entries
Background advances in locationaware devices web mapping technologies mobile cartography Contribution of none professional volunteers in providing and distributing geographic information
Background http://dicomputing.pnl.gov/
Background Events and their patterns represent a higher level of knowledge Highly complex and irregular patterns still can be represented visually for further analysis Efficient planning and management of a complex system.
Background
Objective To extract event related knowledge from OpenStreetMap Retrospective event refers to anomalous user activities, which happened at a time or within a specific period of time. User activity refers to number of contributors plus number of contributions
Case Study 1 Stoppelmarkt in Vechta, Niedersachsen, Germany One of the oldest folk festival in Germany (since 1298). Once a year from Thursday to Tuesday in the week of 15 th August. Approximately 160,000 square meters
Case Study 2 Oktoberfest, in Munich, Bavaria, Germany One of the most famous events in Germany The largest world's folk festival. Mid-September until first week of October. Yearly more than 6 million people attend this fair and celebrate the Oktoberfest.
Case Study 3 The earthquake and tsunami in Japan, 2011, one the most powerful earthquakes in the world The powerful tsunami waves traveled up to 10 km inland Extensive and severe damages to roads, railways and other infra-structures This event caused dramatically changes in the geometry of objects, which are important for a map.
Case Study 3 The earthquake and tsunami in Japan, 2011,
Results (Case Study 1) No detectible indications of unexpected increase in the amount of contributions. Slow changes in the number of contributions and contributors
Results (Case Study 1) Tag Mining: Information is only available in advanced editing modus
Results (Case Study 1) Tag Mining: Objects are constantly displayed on the map
Results (Case Study 1) Tag Mining for Maximilian s Pub: Exploring some relevant information about temporal characteristic of stands and the markets Edited at: 04. Aug 11 05. Aug 11 05. Aug 11 28. Jan 12 Edited by: Duetzer Duetzer Duetzer Nolbelt San Version: 1 2 3 4 In changeset: 8916581 8925849 8925909 10524834 Comment: ------------------- ------------------- ------------------- small changes amenity pub pub pub pub Tags: name Maximilian's Maximilian's Maximilian's Maximilian's opening_days ------------------- ------------------- 11 to 16-Aug-2011 11 to 16-Aug-2011 Coordinates 52.74777, 8.29669 52.74777, 8.29669 52.74777, 8.29669 52.74777, 8.29669
Results (Case Study 2) An upward trend in the number of contributions and contributors No easily detectible indications of unexpected increase in the amount of contributions.
Results (Case Study 2) An upward trend in the number of contributions and contributors The ratio between amount of contributions and the number of contributors decreases over time. New generation of OSM contributors is busy with enhancing and maintenance of OSM maps.
Results (Case Study 2) Tag Mining: Tents not visible in the view layout of the map,
Results (Case Study 2) Tag Mining: The volunteers, identified the start, and end dates of the event, and have hidden the temporal tents
Results (Case Study 2) Tag Mining for Augustiner tent: Altering the hidden tag from yes to no Adapting the date after the event for next year Edited at: 27. Sep 10 28. Sep 10 04. Okt 10 04. Nov 10 01. Jan 11 17. Sep 11 26. Sep 11 07. Nov 11 Edited by: mackerski mackerski mackerski sendelhorst spunsel chan mackerski swus Version: 1 2 3 4 5 6 7 8 In changeset: 5894426 5901895 5954479 6292634 6828639 9326733 9402214 9766624 Comment: ------------- ------------- ------------- hide update dates unhide unhide hide start_date ------------- 01. Sep 10 01. Sep 10 01. Sep 10 17. Sep 11 17. Sep 11 17. Sep 11 22. Sep 12 end_date ------------- 20. Okt 10 20. Okt 10 20. Okt 10 03. Okt 11 03. Okt 11 03. Okt 11 07. Okt 12 hidden ------------- ------------- ------------- yes yes no no yes name Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner Tags: operator Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner Augustiner building beer_tent beer_tent beer_tent ------------- ------------- ------------- beer_tent ------------- seasonal_building ------------- ------------- ------------- ------------- ------------- ------------- ------------- beer_tent temp_building ------------- ------------- ------------- beer_tent beer_tent beer_tent ------------- ------------- note ------------- update update update update update update update The tents were created in September 2010 by Mackerski and have been updated regularly for the next year by different users
Results (Case Study 3) An obvious increase in the number of contributors and number of contributions over a very short space of time Change in the geometry and consequently the semantic information of affected objects
Results (Case Study 3) Edits in February 2011
Results (Case Study 3) Edits in March 2011
Conclusion Finding keys from Social events: OSM is not a tool which citizens would naturally turn to mark an event contributors do not treat this community in the same way that they contribute on user-friendly electronic exchange platforms such as Twitter, Facebook, or Flickr. Activity of contributors (Number of contributors plus Number of edits) for a specific period of time cannot be considered as an appropriate indicator for detecting social/temporary events.
Conclusion Finding keys from Social events: Size or popularity of an event does not have any significant effect on the possibility of event detection in OSM. OSM like other sharing platforms have the possibility to provide the semantic information for the mapped objects by tagging them. Overall there appears to be a lack of event detection and pattern recognition possibility form OSM community for the social and temporary event.
Conclusion OSM deficiencies for social event detection: Users do not have access to the embedded information (e.g. start and end date of events) Mapping of temporary objects for repetitive periodical events affect the quality of OSM for offline users Continuous maintenance is needed to update the map for each period. Many events are not manageable to be presented through hide and unhide mechanisms.
Thank You