21
$\begingroup$

Both on this site and networkwide there are many links to CiteSeerX. it seems that those links no longer work - at least the ones that I checked.

  • Is there some official announcement explaining what happened to the site and whether it is definitively discontinued?
  • Do you happen to know whether the links to this site might start working again in the future?

Are there reasonable ways to fix such links?

  • If the OP included enough information about the paper in the post, it probably doesn't matter whether the link no longer works - knowing at least the author and the title is enough to locate the article.
  • In some cases the link might have been archived in the Wayback Machine. But we will not always be that lucky.
  • The links contain some kind of identifier - but it is unclear how to get further data from that identifier. (See also this question on Academia: How does CiteSeerX's DOI parameter work?.)

I tried to check whether there are some posts where Wayback Machine links have already been added for those links.

Search returns 22 posts with URLs containing *web.archive.org*citeseer* networkwide. (None on this site, 3 on MathOverflow.)

SEDE returns 5 posts on MathOverflow and posts on Mathematics. (Some of those results are false positives.)

Looking specifically at the link from this MathOverflow post, we can see that the Wayback Machine link uses the id 10.1.1.43.3848. The same paper can be now found under pid/c4d9cd7ba35b6d295fe19025eafe4561fe19d75e. So this instance would suggest that CiteSeerX changed the identifiers they're using.

While this could be seen as a networkwide issue, I posted here. One reason is that I do care more about the posts on this site rather than elsewhere. The other things that efforts to fix dead links seem to be rather active on this site. Of course, if more experienced people suggest that Meta Stack Exchange would be a better place for this, I will post there.

$\endgroup$
8
  • 1
    $\begingroup$ I don't know if it is related, but quite recently the following happened to me. I googled with a few buzzwords, and was referred to an article on CiteSeer. The problem was that I couldn't copy the url with a mouse from anywhere. My browser only offered the options of downloading or opening the pdf directly. I thought that may be it was the way Google generated the links. Anyway, I still managed to type in the url by hand, and got the link to work (here in MSE). $\endgroup$ Jan 5, 2023 at 17:30
  • 3
    $\begingroup$ Yeah, they definitely seem to have modified their url routes. I didn't figure out any obvious way to migrate one schema to the other but I'll look some more. $\endgroup$
    – rschwieb
    Jan 5, 2023 at 22:03
  • 4
    $\begingroup$ Oh no... flashbacks to SpringerLink $\endgroup$ Jan 6, 2023 at 2:49
  • 2
    $\begingroup$ There are quite many posts pertinent to this problem in CV. While editing broken links, fortunately, I could find archived links. But this seems to be pricking issue there too. $\endgroup$ Jan 6, 2023 at 4:24
  • 3
    $\begingroup$ Might be worth trying the contact link (csxstatic.ist.psu.edu/contact)? Just FYI for the SEDE link on Math.SE, this post links to https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CDIQFjAA&url=http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.187.737&rep=rep1&type=pdf&ei=7P8bUYmxEIup0AHErIHIAw&usg=AFQjCNH1CGocnVyf24ef4TSkrcwB63r-8g&bvm=bv.42261806,d.dmQ&cad=rja which is a redirect to https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.187.737, and is broken $\endgroup$ Jan 6, 2023 at 7:40
  • 1
    $\begingroup$ In the other link found in the SEDE query, the first and second links have replacements. Currently searching for url:*citeseer* gives 930 hits on Math.SE. $\endgroup$ Jan 6, 2023 at 7:47
  • 1
    $\begingroup$ @CalvinKhor: Putting "Introduction to Cobordism-Weston" into the Citeseer$^X$ search box leads to a PDF Introduction to Cobordism Theory by Tom Weston. So presumably the tags/identifiers used by Citeseer$^X$ have evolved, but much of the content (esp. PDF files) is retained. $\endgroup$
    – hardmath
    Jan 6, 2023 at 22:47
  • 2
    $\begingroup$ Is this a recent problem? I'm not sure if there is a direct correspondence, but every file in the Github page (found from the Academia.SE page) hasn't been touched in ≥3 years $\endgroup$ Jan 7, 2023 at 1:45

3 Answers 3

5
$\begingroup$

About 3 weeks ago, I opened an issue in the Github page of CiteSeerX. On the 20th of January, I got the following response from shauryr (Sorry I didn't check earlier, been busy!)

We are aware of this issue. We are working on a redirection service, which will make the old links work. The update will be out within the next week. Thank you for your patience.

So fingers-crossed, the issue will resolve itself.

(Update Feb 20, 2023) I have tested two links (not a lot, but...) and the issue does indeed seem to have (another editthanks The Amplitwist mostly) disappeared!

$\endgroup$
4
  • 3
    $\begingroup$ (i note it is past the week after I got the response, but I still hope the update will arrive) $\endgroup$ Jan 28, 2023 at 2:52
  • 1
    $\begingroup$ Relevant to the update, there was also a short conversation last week in the chatroom Boulevard of Broken Links. $\endgroup$ Feb 20, 2023 at 7:42
  • 1
    $\begingroup$ @TheAmplitwist ahh I was too optimistic. Thanks :) $\endgroup$ Feb 20, 2023 at 8:36
  • 2
    $\begingroup$ As pointed out in chat, the problem with the two links mentioned above was that they were actually links to citeseer.ist.psu.edu rather than citeseerx.ist.psu.edu. There are a few links to the "old" domain on this site and network-wide. $\endgroup$ Feb 24, 2023 at 7:58
22
$\begingroup$

I sent an email to CiterSeer$^X$ using the form on their contact page, and I'll advise of any response I get:

I'm a user at Math.StackExchange.com involved in maintaining links and triaging reports of breaks.

Recently a large number of URL's in our posts that match citeseerx.ist.psu.edu/viewdoc/summary?doi=... have stopped working. In a number of cases I'm able to find replacement links that are instead of the form citeseerx.ist.psu.edu/pdf/...

I don't see any explicit mention on your site of a reorganization of the site's links, but of course this sort of thing occurs by necessity over time. Would you be able to shed any light on the circumstances with an eye toward the most expeditious way to repair such broken links?

Regards, (etc.)

Update: I had not received a response from my query to CiteSeer$^X$, so (gasp!) I picked up the phone and spoke to someone at Penn State who is working on the problem. He was optimistic that the link redirection will be fixed, maybe as early as tomorrow. At any rate they are aware that with the reorganization of links and identifiers, much was broken.

Here is an issue opened last October in a github project that notes these link breakages and how they fixed them in their own documentation:

Files pointed by the citeseer or citeseerx URLs are no longer reachable. I think the site has updated its routes to the files.

The good news is that there are only 25 places in 17 files to be modified...

For example, the URL in clustering.rst

"A Random Walks View of Spectral Segmentation" http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.33.1501_

should be replaced with https://citeseerx.ist.psu.edu/pdf/84a86a69315e994cfd1e0c7debb86d62d7bd1f44 or https://citeseerx.ist.psu.edu/doc_view/pid/84a86a69315e994cfd1e0c7debb86d62d7bd1f44

They made the changes to URLs in their documentation and closed the issue shortly after (Nov. 2, 2022). This suggests our approach will probably need to be similar, with a certain amount of tedious search and reconstruction of valid links.

I'm going to hold off building a spreadsheet to track the broken links and repairs here and for MathOverflow until the repairs that CiteSeer$^X$ has in progress come into focus. Ideally we can limit the effort to verifying that most links just start working again.

$\endgroup$
5
  • $\begingroup$ I know it's not quite the point of the question, but can't links with DOIs, even if hidden in the URL parameter, be migrated directly and probably more or less automatically to https://doi.org/… (and then people can Google CiteSeer or other links themselves)? $\endgroup$
    – LSpice
    Jan 11, 2023 at 22:29
  • 4
    $\begingroup$ @LSpice: The abbreviation DOI (Digital Object Identifer) does not denote a common value designation between the two sites, although the visual similarities make it tempting to suspect it could be so. The doi.org syntax is apparently evolving as well. $\endgroup$
    – hardmath
    Jan 11, 2023 at 23:51
  • $\begingroup$ Re, thanks—I didn't realise those two were different. As for the second part, although I'm getting off the point—although the doi.org syntax evolves, I think they don't invalidate existing DOI's, right? $\endgroup$
    – LSpice
    Jan 12, 2023 at 3:49
  • 1
    $\begingroup$ @LSpice: I'm guessing doi.org is pursuing a pretty conservative path in terms of preserving the existing identifiers, but their News page is pretty stale (as well as that link I gave you, now two years old -- work in progress). $\endgroup$
    – hardmath
    Jan 12, 2023 at 4:34
  • 3
    $\begingroup$ @LSpice There was a related question on Academia: How does CiteSeerX's DOI parameter work? (Actually, I included the link in this meta question.) $\endgroup$ Jan 17, 2023 at 8:49
10
$\begingroup$

They broke ~17,000 links on Wikipedia with this change.

Someone there has found how to convert old CiteSeerX URLs to the new format:

I found the jackpot, of sorts. A data dump that relates the old CiteSeerX ID to the SHA1. https://archive.org/download/citeseerx-csx_citegraph.2017-03-31/citeseerx_checksums.tsv.gz (expands to 624 MB) contains over 10 million entries of simply (SHA,ID). So if I have 10.1.1.676.1062, I can search the file and I find dc6437569a8a2ddd1c22ef623f8fdd6e74a1b535. Voilà, now I can access the new CiteSeerX website: https://citeseerx.ist.psu.edu/doc_view/pid/dc6437569a8a2ddd1c22ef623f8fdd6e74a1b535

So, here is how to do the conversion:

  • Find the DOI in the old URL (for instance, in http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.43.3848, the DOI is 10.1.1.43.3848)
  • Download and extract https://archive.org/download/citeseerx-csx_citegraph.2017-03-31/citeseerx_checksums.tsv.gz
  • Search lines containing the DOI in the downloaded file
  • The first column of each line is the hash to use in the new URL format. In c4d9cd7ba35b6d295fe19025eafe4561fe19d75e 10.1.1.43.3848 pdf, c4d9cd7ba35b6d295fe19025eafe4561fe19d75e is the hash
  • The new URL is https://citeseerx.ist.psu.edu/pdf/ + the hash: https://citeseerx.ist.psu.edu/pdf/c4d9cd7ba35b6d295fe19025eafe4561fe19d75e
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .