Like everyone else I use Google scholar to search for papers that might match the topic I'm looking for. Compared to the alternatives, it has good coverage and a good full-text database of papers, not just their abstracts, so I get more relevant hits that way. But lately once I find something I want to cite I've been switching to MathSciNet and searching for it again. The reason: MathSciNet has cleaner BibTeX data than what I get from Google scholar or the publisher web sites.


sudoman: The Collection of Computer Science Bibliographies