The many upvots on Gerrit's answer show how the majority of researches currently does search, filter, read scientific journals and books, as the best way, a context-sensitive search(X Algorithm near to Y algorithm near to Z database) is not available in scientific search engines to my knowledge? A lot of time has to be wasted for filtering out uninteresting content.
An google WEB search example would be:
etching AROUND(4) redeposition
Search hits will only show sites where etching is separated from redeposition by not more than 4 other words between those two. While many research papers often contain many keywords in intro and abstract of a paper (for retelling recent history in a field), which makes finding papers in a special context impossible (measuring effect X with method Y on material Z), in principle the AROUND(x) operator gives you this power back.
The problem is, google SCHOLAR hasn't implemented the AROUND operator, therefore I currently use it sometimes in google WEB search when searching papers (unfortunately only titles and abstracts are indexed by Google, but main body content isn't searchable) in a very specific context.
I'm really wondering if there are secret hacks for free (google scholar, scirus...) and commercial (ISI, scopus) search engines giving you the power of context-sensitive search, as it would make a lot of reading, searching, filtering, rating, sharing unnecessary.
My current way to solve this is downloading many highly-cited papers (withoud reading) as gerrit's answer shows and indexing it with my own desktop search engine (Copernic desktop search software) which has a similar operator. It really ticks me off, that a search technology that is available for years is not brought to the place where it would be most useful?! Maybe we should similar to thecostofknowledge just start to build our own search engines? There are some business reasons (traffic, downloads,..) that make the non-implementation of this operator in commercial search engines very reasonable to me...
So, am I missing some secret hacks or search engines that are near to context-sensitive/semantic search?