Curious.Judith

While JCDL doesn’t have a necessary link to geospatial data, i do hope that as time passes we’ll see more. I remember the spatial visualization Perseus Digital Library introduced years ago; with the map APIs and other resources geospatial visualization should be more accessible.

This talk delivered by Rahul Nair about work he and his colleagues have done with a corpus of geolocated and tagged flickr photos crosses a large number of my interests. In very brief, they take the geolocated photos, identify spatial clusters, look for common tag terms, and use those tags to generate place name labels. (One can then browse geolocated Flickr photos with those tags in their interface.)

The textual clustering to come up with labels appeared to be remarkably successful compared to the automatic cluster labeling experience I had using the Recommind engine on bibliographic data. I do believe that the fact that the photos were intentionally geolocated (as compared to auto-captured metadata in a GPS enabled camera) helps to improve the results. Users don’t incidentally but intentionally associate that image with place. I do wonder if this demonstrates that users (at least these who are willing to geolocate their images) are willing to carefully construct their tags to have general location tags as well as more specific. The apparent success is quite satisfying.

Another piece they touched on was the dynamic cartography or neocartography principles in play. Most important was that the labels were scale sensitive. Cities and regions at small scale; neighborhoods an particular features at larger scales. Rahul noted that in user testing there were some labels users identified as “missing,” which they believe the occlusion algorithm removed. In manual cartography, one is able to make subtle adjustments to make those placement decisions. ArcGIS has rules for label overlap — are there other, dynamic ways of identifying multiple labels? Google Earth’s dynamic method of exploding points out is an attempt to do this on mouse focus, but it hasn’t seemed as successful as one might wish.

For the neogeography crowd — tagmaps.research.yahoo.com/ is offering up an API so, given a bounding box, one can receive back the tags for that region. And for wherecamp attendees — it seems that they use a geotude-like tiling, using equal lat-long tiles at fourteen levels of zoom to display the tags.

Rahul spoke about applying their methods to other geo data like geoRSS feeds (and one wonders whether KML might be in mind). I’ll be interested in seeing how that turns out.

(Later in the session was a visualization of citations regarding the Sloan Sky Survey and the “cosmospatial” map view. I’m looking to see where it lives on line….)

World Explorer: Visualizing Aggregate Data from Unstructured Text in Geo-Referenced Collections (Flickr oriented)
Shane Ahern, Mor Naaman, Rahul Nair and Jeannie Yang
Categorization and Analysis of Text in Computer Mediated Communication Archives using Visualization
Ahmed Abbasi and Hsinchun Chen
Delineating the Citation Impact of Scientific Discoveries
Chaomei Chen, Jian Zhang, Weizhong Zhu and Michael Vogeley

JCDL2007

People who looked at this item also looked at…

• Contact Us
• 3×5 envelopes for your hipster PDA system
• FaceBook introduces me to the Gwen+HP+Fan book
• Remote Imaging: Northern New Mexico and LANDSAT analysis
• Grey Brother

Related items

• Cooperative Cataloging: from music to location
• Friday morning sessions
• BBC’s The Box Project
• GPS+Camera+a Mac+HoudahGeo = Wow!
• More identity games — and Plaxo

Tags: ArcGIS, Categorization, Chaomei Chen Jian Zhang Weizhong Zhu, Citation Impact, cities, Computer Mediated Communication Archives, Delineating, Flickr, Geo-Referenced Collections Flickr, geolocated, geoRSS, Google Earth's, gps, Hsinchun Chen, I'll, I'm, jcdl, Jeannie Yang, KML, labels, Michael Vogeley, Perseus Digital Library, photos, Rahul, Rahul Nair, Recommind, Scientific Discoveries, Shane Ahern Mor Naaman Rahul Nair, Sloan Sky Survey, Tagged, Text, Unstructured Text, Users, Visualization, Visualizing Aggregate Data, While JCDL, World Explorer

This entry was posted on Wednesday, June 20th, 2007 at 12:03 pm and is filed under GPS and GIS, TunaBreath. 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.