Lee Bryant of Headshift talks about five case studies.

1. In an experiment, users get to tag stories on the BBC news page and see other people’s tags. (Stowe blogged about this here.)

2. Local aggregation, pulling in news, blogs, links, photos and government information. Text analysis derives common themes. Show the keywords by frequency so people can navigate. Make everything comment-able.

3. Building shared meaning with tags — cluster and search on themes.

4. Negotiating language — learning from unstructured data. They asked for user-driven feedback on health services based on unstructured stories. They use text analysis to identify matches with formal taxonomies. If the user says that’s wrong, the system learns. So, they create “user-generated tag clouds.”

5. Where do we go from here? His company is trying to escape from the “taxonomy deadzone” by providing a lightweight social software interface.

Q: (Dina Mehta): How do you talk a company into taking this bottom up approach.

A: Point out what isn’t working. Show them how they can build on top of what they have.

[Excellent talk.]

Categories: Uncategorized dw