Dimensional tagging approaching collective intelligence
At last week's Future of Web Apps conference, Bradley Horowitz of Yahoo presented an example of the power of tagging in flickr, a photo sharing website. I have replicated the example which was a search for photos tagged with geographic coordinates and the phrase route66. When plotted on a map, the results cluster nicely along the legendary highway Route 66. The map represents an example of collective intelligence: No single person had to travel the entire length of the highway to produce it.
The example shows how large collections of data can reveal patterns not evident at the item level. Could flickr's innocuous hobbyist photographers help us reveal, say, crime hot spots in London? Maybe with time; so far there are less than 100 data points:
The approach to reveal collective intelligence is illustrated with photos in this example, but there are many potential other applications: Customer records, documents, call centre entries, laboratory measurements and most kinds of transaction records. The central requirement is that people are free to apply tags to the data. Any tags. If the only option is relating items to a taxonomy, the data set will only answer yesterday's questions (for which the taxonomy was designed).
Working with information in this manner means dealing with a new type of complexity which springs from the fact that the tags originate in individual human minds. The Route 66 example does not result in a perfect tracing of the US Highway - somebody seems to have gotten lost in Arkansas and Utah. And there is a semantic angle as well, aptly illustrated by the fact that one flickr photographer has added a crime tag to the photo of ducks and added the coordinates for Kentish Town, a part of London. This is where statistics come in with algorithms for identifying and dealing with outliers.
What data collections do you have or are being built where tagging can be exploited to reveal patterns?
- Tag clouds with a silver dollar lining - about tags, again with flickr as an example
- New geography 101 - about geotagging and GPS