She talks about novelty analysis, using the "Difference Engine" they've .... They have a prototype called News Junkie that pesonalizes news using onfromation novelty:
This is then followed by ..."he talks about noisy channel information finding. For example, in spelling correction the system tries to figure out what the original (intended) spelling was based on the misspellings introduced by a "noisy channel." We're moving from "doe sthe page contain the query terms" to "Does the page satisfy the information need."
I see a complete behaviour methods being captured, used as part of aHerustic Model to sustain branched patterns growth, along with Confidenece Rating per method. It reminds me our oldf GEOIND project.
On this thought flick'r , I have a blog here !! ( a little bit more deeper, I think).