Groups keyboard shortcuts have been updated
Dismiss
See shortcuts

Rounding counts of results in web interface

49 views
Skip to first unread message

Schares, Eric J [LIB]

unread,
Feb 12, 2025, 5:56:14 PMFeb 12
to openalex-...@googlegroups.com

Hi all,

 

I noticed in the OpenAlex web interface today that the count of results are getting rounded off and not displaying exact numbers. For example, type:article and years: 2019-2023 gives 32,380,000 in the web interface, but the same API call gives 32,380,779.

 

It seems there has been a change in the way numbers are reported in the web interface, possibly related to Kyle's recent Usage Updates and edits intended to reduce strain on the servers. I would like to request the web interface revert back to showing exact numbers for a query, as I use these values quite often. Assuming that anyone who wants the exact values will use the API (or know how to use it) is a mistake.

 

Thank you,

Eric

 

Eric Schares

Engineering & Collection Analysis Librarian

Iowa State University

150 Parks Library | Ames, IA  50011-2102

https://eschares.github.io/

https://orcid.org/0000-0002-6292-8221

webinterface.PNG
API_call.PNG

Gabor Schubert

unread,
Feb 13, 2025, 3:40:48 AMFeb 13
to OpenAlex Community
It seems like values under 10000 are still reported without rounding on the web interface. But values over 10000 are rounded off with different thresholds.
Gabor Schubert
Stockholm University Library, Sweden

rounding.jpg

Bianca Kramer

unread,
Feb 13, 2025, 9:51:39 AMFeb 13
to Schares, Eric J [LIB], openalex-...@googlegroups.com
Hi Eric, all,

It's not new behaviour if I'm correct - I remember it being this way for a while already (at least some months) ? 

It would be interesting to know if it's really done for lessening strain on servers, as I fully agree with you that showing exact numbers in the web UI would be far preferable from a user perspective. 

Regarding the API, the fact that the API-url is readily available does help - it would help even more if it would be a clickable link. 
(but still would not be a replacement for the webUI for many users)

kind regards, Bianca

Bianca Kramer
Sesame Open Science





Op wo 12 feb 2025 om 23:56 schreef 'Schares, Eric J [LIB]' via OpenAlex Community <openalex-...@googlegroups.com>:
--
You received this message because you are subscribed to the Google Groups "OpenAlex Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openalex-commun...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openalex-community/PH0PR04MB822635A42B2CD36EDBB2A505C7FC2%40PH0PR04MB8226.namprd04.prod.outlook.com.

Eric Jeangirard

unread,
Feb 13, 2025, 11:19:22 AMFeb 13
to Bianca Kramer, Schares, Eric J [LIB], openalex-...@googlegroups.com
Hi !

The web UI does an API call that gets the exact figure, and then the UI does this rounding in javascript (cf screenshot below). So it seems to be an UI choice to do this rounding. Ideally, we could open an issue on https://github.com/ourresearch/openalex-gui but that does not seem to be doable.

image.png

Schares, Eric J [LIB]

unread,
Feb 13, 2025, 12:12:48 PMFeb 13
to Eric Jeangirard, Bianca Kramer, openalex-...@googlegroups.com

Thanks all. Apparently, I haven’t been very observant! I also sent this question to OpenAlex support and got a response that the web UI uses the toPrecision filter

to round to 4 significant digits, and it has for a few months now. That’s why Gabor saw values 9,999 and under not getting rounded, but those over 10k had a 0 as their last digit.

 

This was explained as not being done for performance reasons, but to avoid the illusion of precision where it doesn’t exist. The exact count of records returned by a filter fluctuates, and reporting an exact number can give users a false impression of certainty.

 

Thanks,

Eric

 

Eric Schares

Engineering & Collection Analysis Librarian

Iowa State University

150 Parks Library | Ames, IA  50011-2102

https://eschares.github.io/

https://orcid.org/0000-0002-6292-8221

 

From: Eric Jeangirard <ejean...@gmail.com>
Sent: Thursday, February 13, 2025 10:19 AM
To: Bianca Kramer <bianca...@gmail.com>
Cc: Schares, Eric J [LIB] <esch...@iastate.edu>; openalex-...@googlegroups.com
Subject: Re: [openalex-community-group] Rounding counts of results in web interface

 

Hi !

 

The web UI does an API call that gets the exact figure, and then the UI does this rounding in javascript (cf screenshot below). So it seems to be an UI choice to do this rounding. Ideally, we could open an issue on https://github.com/ourresearch/openalex-gui but that does not seem to be doable.

 

wordslikethis

unread,
Feb 15, 2025, 7:45:16 AMFeb 15
to OpenAlex Community
Thanks Eric for passing my answer along to the group. I'm reproducing it here below in its entirety in case it's helpful to anyone:

Thanks for the question! It could be that you are just now noticing it, but actually the UI has rounded the results count (and most other counts) from the start. Or to be more precise, the UI reports counts down to 4 to significant figures.

 You can check the code of where the results count is inserted into the template here, and can see it uses the toPrecision filter which is defined here. If you look through the commit history you can see it's not new, so (cue LeVar Burton voice) you don't have to take my word for it. 📖🌈😄

Why do we do this? The same reasons scientists do it: because reporting results to the last digit, regardless of the size of the number, can give users the illusion of precision where it doesn't exist.

It takes time for OpenAlex to process works and aggregate data about works, so the counts we report are always wiggling around a bit, and they generally differ a bit depending on how you aggregate them. I look at this the same as how telescopes images suffer from atmospheric abortions; the accuracy of the instrument is never perfect. Of course, precision can improve, and I have good news there: our early experiences with the rewrite, launching early q2, indicate we've improved precision a lot yay. 

Anyway, we assume technical users of the API understand the wiggles and will discount tiny fluctuations in reported counts; however we report those counts to web users in a way that underscores the inherent limits of precision in our system. I believe this is why general-purpose search engines like Google do the same thing with their results counts.  

I totally get there are lots of users who want full-precision counts but don't want to use the API. So we're launching a new analytics UI in q2 and that will report full-precision counts as well.

Thanks again for getting in touch with the question!
Best,
Jason 

Reply all
Reply to author
Forward
0 new messages