Talk:Visual Analytics for Sustainability and Climate Change/Tool

Some initial comments from Flo

Based on [1], how are « Instances » updated at the moment ? Is that fixed ? If fixed, how can new instances added to Wikidata be added to the vizualization system ?

The visualisations should not be based only on the English characteristics. There are lot's of "translanguage" information that could be interesting visually. For example... seeing the difference in size between French and English ? (if bigger in French... suggest translation from French to English is an option). Or seeing the differences in evaluation (if better assessment in French... might suggest the French article could inspire the English editors). Or huge pageviews in Spanish, but super low in French... what does that suggest ? Etc.

« Number of editors" and "number of revisions", which provides a measure of "interest" around the topic and likelyhood to progress", are very important elements to draw from as well.

It also VERY interesting to visualise the articles that were written a "long time ago", but basically have been left "to rot" since then. An important sign of possible "oudated content". But this would probably be harder to visually represent (maybe somethink such as xx % of the bytes added were added over 10 years ago might be an approach).

Comments from Nicole

• Some guidance in the clustering task will probably help: Can articles only belong to one cluster or none/multiple? How many clusters do you roughly aim for?
• I think "instance" will be key in clustering. Just to keep in mind: It makes also sense to then feed back information into Wikidata (i.e. tag Wikidata objects with the correct instance if they currently have none) - this might make the work more reproducible in the future.
• Some articles seem to be not missing in the visualisation - why is that? For example, if I select instance "climate change by country or territory (13)" only two articles are visualised (climate change in India and climate change in South Africa).
  • The answer is to change the minimum page views (by default at 100). But many entries exist, that are clearly not attached to the Wikidata instance Anthere (talk)
• Very much agree with Florence: Something like number of revisions or average daily revisions would be an important measure - also to better understand whether these were constantly growing articles or articles that referred to current events/were written in a burst (and are now rotting).

Issues (feedback)

So far I find these are some issues to fix with this tool (let me know if I should post these elsewhere, e.g. as issues to a github repo):

• The tool is named "Visual Analytics for Sustainability and Climate Change" but it's general purpose and could be used for all sorts of things so I'd suggest to rename it to sth broad and descriptive like "Wikipedia visualization tool"
• Could you please enable to see the labels of the articles directly in the visualization without having to hover over them? This is also useful/needed for image exports (which can be a main way this tool gets used). For example, articles with over 500 pageviews and less than 6 images
• Why is the disambig page "English" included in the default visualization? Seems like there's some bug.
• Is there any way to get a permalink to the configured visualization? Preferably even with the same zoom state so the same area is visible. As with the labels displayed in image exports this would enable sharing.
• When limiting files via Quality assessments or any other way, please enable hiding the filtered-away dots instead of just greying them out. They clutter the chart, make things hard to see and especially make it difficult to hover over the right dot.
• When limiting files via Quality assessments or any other way, please add a button to zoom to the scale at which all dots can be seen if not automatically zooming to it.

@G.prof, Iopensa, and Cetirizina: Prototyperspective (talk) 23:30, 17 March 2026 (UTC)

For those points, I moved them to Visual Analytics for Sustainability and Climate Change/Tool and I added comments for some. Anthere (talk) 19:33, 1 April 2026 (UTC)

Additionally I'd like to know how to add a new topic. I could not find a guide here. Seems like I need two CSV files, is that right? If so, how to create these CSV files? Prototyperspective (talk) 23:37, 17 March 2026 (UTC)

So I did myself for a test, twice. The first time, there was a sample to show how the format of the file should looks like. And when I did it the second time... the sample was gone. I had to do trials and errors... I agree that the sample or the explanation are needed.

That being said, my second trial was accepted as « data » but the system did not analyse it. I have to ask...

• This is my first batch of articles added : https://impact-visualizer.wmcloud.org/topics/15
• This is my second batch : https://impact-visualizer.wmcloud.org/topics/38 For the users csv, this is fairly straightforward... simply usernames, in one column, one user per line For the articles csv, same... simply titles, in one column, one title per line That being said... the VIZ TOOL currently does not run on examples set by users (I can be proved wrong on this one). Afaik, It only works on the set of articles currently being tested as part of the VIZWP research project. Of course, the goal, once the 3 selected types of visualizations have been implemented, is to connect it to the other selections set up by users. Anthere (talk) 19:46, 1 April 2026 (UTC)

Thanks for this info. It would be nice if G.prof, Iopensa, or Cetirizina could help out too. Now I've accidentally imported a long list of articles in a csv into the users. I clicked on edit topic (while it was still loading 'users' which actually were article titles) and set the correct csv into the articles and a csv with just my username into the users and clicked the blue submit query button but now at the topics page both are stuck at 0%. Can a topic with false csv imported incompletely earlier be fixed like that? And what goes into the usernames CSV? And I can see the article sample csv and it has multiple columns: what are the second and third columns for (categories?)? Prototyperspective (talk) 21:56, 1 April 2026 (UTC)

Set up a new topic and imported the same csv files there (this time correct right away) and it's also at 0%. Maybe the tool has some issues now that prevent loading any csv. Prototyperspective (talk) 21:59, 1 April 2026 (UTC)

It loaded now: https://impact-visualizer.wmcloud.org/topics/37 (topic: science)

Currently it uses only WikiProject ratings. A problem with that is that many articles don't have these set so probably it would be good to add pages in categories in addition. Problems with that are that

1. petscan only works down to a certain level
2. that there can be many offtopic articles due to unfixed miscategorizations (W393: A way to see why a file is somewhere underneath a specific category (tool to show cat-path) would help with this by the way)
3. it needs deduplication of article titles (I think I can solve this one)

Of course, ideally, all articles that are subject of a certain topic should also be tagged with that topic's WikiProject (in this case 'Science') so optimally this would be addressed via some method to add the WikiProject tags to the talk pages of all the articles that miss it automatically via script/bot. Prototyperspective (talk) 15:29, 2 April 2026 (UTC)

Further issues I noticed as this is being loaded:

• "Einstein-Podolsky-Rosen paradox" is "Einstein–Podolsky–Rosen_paradox" in the CSV file; that could be due to issues in the petscan csv export (ideally that issue should be fixed there) – now I don't know if the tool loaded it correctly: apparently there are only 3 articles "not found" so far...is there a way to see which? And is there a way to correct these or add additional articles, eg via providing an additional csv files?
• Looks like I have some text in the second column too despite that I thought I had removed it. Will this cause any inaccuracies/issues and can it be corrected afterwards?
• The former may be the reason why ATA 100 is included despite that it doesn't have the WikiProject tag and is not in the CSV files I imported. If not, why is it included?
• In the search, when I enter "ATA 100" to find that article I found earlier again, I can't find it. Please make the search work so it highlights/opens exactly the matched articles – eg via showing a list or autocomplete for the included items matching the search which if clicked are opened + highlighted. When I export as CSV I can see that "ATA 100" is included (again despite that it's not in the import CSV). However, I can see that in the CSV from petscan of articles to import, there is "ATA Spec 100" with "iSpec_2200" in the second column. So I thought maybe this is due to a delimiter in the title or sth. And now I found part of the cause: the article it can't deal with is redirect ATA Spec 100/iSpec 2200 – the slash must be why petscan had put this in a separate column in the csv export (should be fixed at petscan ideally). However the tool should not import nonexisting articles. Additionally, Wikipedia needs to clean up WikiProject tags on redirects – here the tag is at en:Talk:ATA Spec 100/iSpec 2200.

Prototyperspective (talk) 22:44, 2 April 2026 (UTC)

I can find the ATA 100 page when I enter it. I think I'll delete the topic and create it anew using a new petscan export where I don't let it export to CSV directly since that causes issues with the delimiter. Einstein-Podolsky-Rosen paradox was loaded correctly. Prototyperspective (talk) 18:29, 9 April 2026 (UTC)

Does somebody know what the users imported is supposed to be for or is that for features that are coming later? Prototyperspective (talk) 23:06, 9 April 2026 (UTC)

Made this thread about the issue of WikiProject tags on talk pages of redirects. I hope the label display directly in the chart will will be added (default disabled). Also useful would be not just having sort options but also filters – e.g. I'd like to filter the displayed items so that only articles with 1 or 0 images are displayed. Additionally, it would be great if one could signify and filter whether the article has a linked Commons category; using both of these things in combination would make this tool very useful for adding media to articles that lack them. Prototyperspective (talk) 14:11, 16 April 2026 (UTC)

VIZWP : visualizing Wikipedia articles

Hello !

There are new elements to look for in the VIZWP tool project.

Two of the features selected during phase I were already developed and published (one fully operational, another one mostly operational). We will have to wait for the next bricks :)
You may test the tool on the dataset we originally created for this purpose (most important articles related to climate and climate change) OR you can create and import your own dataset using the new topic-builder tool.

If you want to test the tool on your side, I invite you to test it here and then drop comments, bugs, suggestions there. Your feedback is very important to us.

Another option is to join us at one of the dedicated events coming soon for a demo and a chat. Check out here : https://meta.wikimedia.org/wiki/Visual_Analytics_for_Sustainability_and_Climate_Change/Community#2026
Next chat is planned on June 6th. Register here: Event:VIZWP_05_06_2026

Thanks. Anthere (talk) 08:39, 27 May 2026 (UTC)

Related tools

• https://gap-atlas.toolforge.org

Anthere (talk) 13:39, 26 June 2026 (UTC)

Tool feedback

With respect to: https://impact-visualizer.wmcloud.org/topics/23

• It's not immediately obvious that the boxes presenting data are themselves clickable objects that surface more information. I found this out more or less by chance and discovered there was a lot more to the report than I initially thought.
• The report presents a lot of numbers while not doing much to ground them in any context or tell me what the significance of these numbers are. There are 994 climate change subjects. Okay? Is that a lot? There are over 1.3 million revisions. And nearly 5 million words written. At some point the numbers become so large they don't mean anything comprehensible. I don't know what story you are trying to tell with these numbers, other than "Wikipedia is edited a lot."
• Seeing when each article was created is interesting. It reflects the trends of Wikipedia as a whole. I assume English is intended here (other parts of the report split out information for different wikis).
• I think the interesting insight here is that even though the articles were mostly created in the late 2000s (as with most of English Wikipedia), the word count only really started to expand around 2019, consistent with an overall bump of interest in climate change around 2019, reflecting increased interest around the world.
• The "article analytics over chosen focus period" is virtually useless. I don't know why the X-axis is an alphabetical listing of pages. One of the pages in the report is the current US president whose page receives orders of magnitude more views. And fitting points for 994 different articles in the width of a laptop screen means they are too compressed together.
  • There are different filters to toggle with what is being seen. For example, the American president can be « skipped » by putting a max of 10 000 on the page views...
• Predicted article quality is probably the most interesting part of the report – it's an opportunity to surface where the outstanding gaps are and opportunities to improve. However, to get information about what those articles actually are, you have to click on a different report, giving me the initial impression that this information wasn't surfaced at all. The visualization also doesn't make sense. Your goal is to depict improvement of quality over time, but you articulated it as a burndown chart, which makes sense if you were trying to measure a decrease in missing articles over time. Which you are, but you are already trying to measure an increase of articles in other classifications. So the size of each blob gets bigger while the whole thing slopes downward. It's a confusing visual metaphor. And why are the two extreme ends of the spectrum represented by gray/white and the others by shades of blue? If somehow all 994 articles became featured articles, is the chart supposed to be a white void?
• The single most useful part of the report is the grid of subjects and language-by-language coverage, and it's buried in the interface, requiring you to click on two different things. There is also no way to discern based on priority, which I think is a bigger problem with how the report is generated. With close to 1,000 different climate change related subjects, as a Wikipedia editor I would want to know what the priorities for translation are. And WikiProject Climate Change (on English Wikipedia) does provide this priority information. I also want to be able to view more than ten subjects at a time.
  • noted on the « more than 10 » items. I also agree on the fact this panel is buried at the moment... Regarding the priorities for translation... it is not the job of a tool to tell the user what are the priorities... this is a human decision really...

Harej (talk) 19:52, 3 July 2026 (UTC)

Thanks for the feedback Harej !

I put inline a few notes. Anthere (talk) 11:49, 8 July 2026 (UTC)

Notes after the Glam Call

1. Languages for the general tool. Only a few are available; Why no Bulgarian? https://impact-visualizer.wmcloud.org/my-topics/new
   1. Comment: I think this is related to the WhoWroteThat dataset not being enabled for all languages yet. See https://phabricator.wikimedia.org/T372340
2. The little graphic top right (the vertical bars, blue/purple/green)... what do they represent ? the graphic is the same for all sets, but does that mean something specific ? If not, why bars of several heights ? Decoration ?
3. Display (axis), in the short list of elements to be displayed: centrality of the article, number of incoming and outgoing links will be useful

Credits: we could add a link to the source code

I was wondering how we could show the source code more prominently. Maybe from the credits page we could add a link "Source code" leading here:

https://github.com/WikiEducationFoundation/impact-visualizer

What do you think about? or, a footer could be introduced. Hoping to attract more code contributions :) Thanks ValerioBozz (talk) 11:28, 5 September 2026 (UTC)

WikiWho: could be abandoned

Thanks again for this tool,

I noticed that the visual tool also uses WikiWho, that seems abandonware since 7+ years, with very low support of Wikipedia languages (and with its current website parked with commercial spam).

So, hoping to be useful, I would recommend to abandon that dependency when possible, to simplify your future developments. Or, become a new maintainer of WikiWho :P :) Thanks again ValerioBozz (talk) 11:49, 5 September 2026 (UTC)