Cite Seer

Although lacking the full search capabilities of Google, CiteSeer's narrower focus means that most results returned are research papers, instead of commercial or personal web sites, which tend to be of minimal value to CS researchers.

CiteSeer discovers papers in two ways. The first is the simplest: you upload a paper to it. CiteSeer allow you to submit URLs of papers to add to its collection. CiteSeer process the files and try to add the paper to their collection. This processing is not perfect, as I've tried submitting one of my papers twice, and it's still not up. Each paper ("document") that is processed is made available in multiple formats (PostScript, PDF, etc.), as well as recording the original URL.

The second way CiteSeer discovers papers is from the bibliography of papers in its collection. CiteSeer process the bibliography to produce a list of "citations" (papers CiteSeer knows exist, because other people referenced them, but it does not have the document itself) that can also be searched.

CiteSeer provides a rank value of both citations and documents, based on how many other papers cite them. The ranking algorithm takes into account self-references (references by one paper of an author's in another paper by the same author), how the reference is done, and the age of the refering documents. Thus, you can see not only who else is referencing, say, your papers, but also "grade" the impact of your papers.

CiteSeer also provides some interesting, although only questionably truly valuable, evaluation of the entire database. One example is the average reference rank of the papers from a given conference (OSDI is highest right now).

Although CiteSeer's focus is on Computer Science, quite a few Mathametics and Engineering papers are in the index because some computer scientist found them useful.

CiteSeer's URL is http://citeseer.nj.nec.com/cs