2 ms·

Is there a reason to think this would be better than just computing the pagerank of every scientific paper based on the citation networks? (Optionally, use sign

Is there a reason to think this would be better than just computing the pagerank of every scientific paper based on the citation networks? (Optionally, use signals to predict future pagerank of recently published papers.) I would expect that using the citation data from published papers would be more valuable, in the same way that revealed preferences are more valuable than polls due to the cost. While Google Scholar implicitly uses pagerank to rank academic papers, I've put together a tool that explicitly shows you pagerank for each paper, available at [1]. Currently, this contains only pubmed, which means that it does leave out a significant set of physics/computer science papers, and also doesn't yet capture preprint archives. 1 = https://pubrank.carbocation.com/ https://pubrank.carbocation.com/

SubiculumCode 9y ago
To discuss how a paper is wrong you have to first cite the paper. The need to discuss bad papers can arise for a number of reasons. So some papers cite rank != quality.
- carbocation 9y ago
  True. Pagerank is "importance" rather than "quality". But as your network of recommenders becomes larger and starts to approximate the scientific community as a whole, I'd imagine that there will be convergence. If not, that would be a welcome and (to me) surprising result.
SubiculumCode 9y ago
Curious. Just looked up my papers, most published in the last few years. How does the pubrank take in account time, or does it?
- carbocation 9y ago
  Pubrank currently is just pagerank based on the citation network. The common way to take time into account is to compute, for example, a "pagerank at 2 years". There are other ways to adjust for time (from linear modeling to more complex approaches) that I'd like to explore. On average, for reasons that are not clear to me but which emerge from the data, pagerank seems to stabilize around 8 years after publication, for most publications.

SubiculumCode 9y ago

To discuss how a paper is wrong you have to first cite the paper. The need to discuss bad papers can arise for a number of reasons. So some papers cite rank != quality.

carbocation 9y ago
True. Pagerank is "importance" rather than "quality". But as your network of recommenders becomes larger and starts to approximate the scientific community as a whole, I'd imagine that there will be convergence. If not, that would be a welcome and (to me) surprising result.

SubiculumCode 9y ago

Curious. Just looked up my papers, most published in the last few years. How does the pubrank take in account time, or does it?

carbocation 9y ago
Pubrank currently is just pagerank based on the citation network. The common way to take time into account is to compute, for example, a "pagerank at 2 years". There are other ways to adjust for time (from linear modeling to more complex approaches) that I'd like to explore. On average, for reasons that are not clear to me but which emerge from the data, pagerank seems to stabilize around 8 years after publication, for most publications.