Just a small update today – the news aggregator I built a few weeks ago can now retrieve tweets related to each news cluster (just click the small bird ).
All I’m doing is taking the three most important words in each cluster (generated via tf-idf weighting) to build a search query that I use to search Twitter by making a YQL client-side call.
I’m also experimenting with OpenCalais to get a richer set of semantic metadata. I’m planning on using the additional metadata to allow browsing by topic/event type, people/organizations involved, and location. Although the metadata isn’t being used for anything useful right now, you can see some of it in the markup (span class=”meta”).