Very Impressive! I've been working on a number of 'content extraction' tools - cQuery.com and webXtract (experimental), but maybe you would find them a useful way to extract news and headlines from hundreds of online sources.
Here is an experiment that uses 'profiles' as a basis for content extraction: