CiteSeer is a
web-based service available at
http://citeseer.ist.psu.edu/cs. CiteSeer contains a database of many, many
research papers that can be searched for papers relevant to whatever topic you are interested in. The database primarily consists of
computer science papers, although relevant
mathematics papers can also be found.
CiteSeer began at the
NEC Research Institute and later moved to Penn State.
CiteSeer is an invaluable resource to anyone doing research in computer science. It processes the PostScript and PDF documents and extracts the text. This means that not only does it know about the documents in its database, but it also knows about documents referred to by documents in its database, a large set indeed. Thus, you can search through thousands of documents (either in the header information or the entire document) and citations in a matter of seconds. If you search for a document by its full title, then any match you get is almost assuredly the document for which you were searching. A similar search in Google or another search engine usually returns many unuseful links that must be sifted-through to get the real document.
CiteSeer can understand TeX files and extract the document's text, but appears to have difficulty processing documents created by other tools. I have been unable to get CiteSeer to process a troff-produced paper of mine. I have not tried Microsoft Word, since I have no want to deal with writing a paper in MS Word, so I cannot comment on PostScript and PDF files created by it. CiteSeer is sometimes confused about the header information, especially author's names. CiteSeer provides a mechanism to submit corrections to this information when the occassional error does arise.
Another task for which CiteSeer is well-suited is ego surfing. You can check how often each of your papers are being referenced, and who is referring to them, giving you a notion of the impact each paper is having.
A draw-back of CiteSeer is that it is often over-loaded. The operators attempt to compensate for this by limiting queries, but it is frustrating to have to continually resubmit your queries when this happens. However, it is free (including advertisement-free), so it is difficult to fault the service.
CiteSeer is a powerful tool that I use regularly and find immensely valuable. It's only real competitor is Google Scholar, which, at least in its current form, is not nearly as useful.