Hacker News Reader

Comment by paulsutter | Hacker News Reader

The current peer review process requires expert supervision (qualified reviewers need to be selected, etc), so a charge of GBP 2,000 seems reasonable to me.

In time we can hope for an HN-like system where publications are ranked immediately on publication. The algorithm would have to recognize the expertise of authorities when assigning the value of each vote. When this is possible, that cost should dissapear. But until such a system has been proven out, peer review should remain and a charge of GBP 2000 is reasonable.

View on HN

1) GBP 2,000 still seems steep to me. Assuming I'll have to pay an scientific editor (who selects reviewers, forwards manuscripts and decision revisions, etc) GBP 100 per hour, he could spend 10 work hours on the article -- total cost: GBP 1,000. I realize that there's a large variance on time required though, depending on how referees and authors agree, but 10 hours seems at the upper end of the spectrum. Of course, if we'll have to pay a typesetter + infrastructure, GBP 2,000 might be where we'll end up. However, I have the impression that these proposals offload all typesetting to the authors.

2) Can you enlighten me (or suggest reading) on how HN ranks submissions? I've had discussions about some sort of ranking of papers, but we generally concluded that the sheer amount of papers published daily would make it tough to get enough (>10?) votes for being somewhat representative I imagine (for example, astro-ph on arxiv sees 50-100 submissions daily). But as you suggest, some algorithm doing preliminary ranking might help this (although it also spells danger).

To rank academic papers you'd need a way to establish certain people as authorities within each field or area, and make their vote stronger. This authority ranking would surely have expert input initially, but I would expect that Citeseer and Google Scholar may be good places to look for inspiration on ways to crowdsource authority rankings.

For starters look at the Page Rank paper (named after Larry Page, not web page). Academic papers need something richer than page rank, but since tens of billion of web pages can be ranked for arbitrary queries, the volume of academic papers should be fime.

Ranking algorithms are one of my favorite subjects, and they are as challenging as they are important. "Search" has little to do with actual searching, it's really a ranking problem. Google's search was the first good ranking scheme for internet search, and their "I feel lucky" button was a way to show that off.

Happy to discuss by email.

Reply on news.ycombinator.com