Interesting paper by several people at Google, including Sergey Brin (co-founder). It describes research work to match text news articles to TV news reports. Clever idea.
It uses the text in the close captions and tries to find news articles on the same event or topic. It's not a recommendation system -- they're finding other writeups of the same story, not different but relevant stories -- but it's still worth a peek.
On a related note, according to Sergey Brin's amusingly outdated resume, he once prototyped a movie recommendation system. Small world.
Update: Nathan Weinberg calls this "GoogleTV" and discusses the implications.
Update: Gary Price lists a bunch of tools that provide keyword search over close caption text (although none of them seem to do it implicitly, which is the focus of the Google paper). And Danny Sullivan talked about blending search and TV a couple weeks ago.