Paper rankings are becoming increasingly important to those in the scientific community, who use these measures to get an overall idea of the impact of their research. More importantly, their use is widespread on Web sites that use citations as a reference for ranking scientific work. As a result, institutions and academia are getting more and more interested in these measures. One of the most widespread measures is the citation count. For example, the h-index is based on the number of citations received. Originally suggested by J. E. Hirsch, this formula ranks the scientific productivity of researchers. Other measures attempt to measure the impact of papers.

Focused on the second goal, this paper provides an intuitive and easy-to-follow description of the main algorithms used to rank papers, including PageRank, CiteRank, and YetRank. The authors introduce a new ranking algorithm, NewRank, which aims to overcome the main drawbacks of CiteRank.

The authors begin with a thorough overview of the architecture of the system they built to conduct the experimental evaluation. Lacking a formal repository that could be used to make comparisons among algorithms, the authors retrieve the whole citation graph from Microsoft Academic Search. Given that these computations are lengthy, the study is restricted to the computer science domain. The algorithms mentioned above are also compared with the simple citation count. Surprisingly, the citation count performs very well when compared to the other ranking measures. Also, NewRank improves on CiteRank and shows promising performance over the other ranking methods.