User talk:MGerlach (WMF)/covid related pages reading sessions

Feedback[edit]

We would appreciate feedback related to:

Does the data feel useful? How would you use it?
Are there topics that seem well out of scope?
If we were to publish data like this in the future, what kinds of information would be helpful for you?

Some initial observations[edit]

Thanks for performing the analysis and sharing the initial results here, which I think have the potential to be quite useful, both for the current pandemic (especially in conjunction with the closely related to COVID-19 corpus) and as groundwork for covering future events.

I would like to be able to play a bit with the data and code, e.g. the parameters that you used. For instance, I think the variant with the virus that you mention to have consciously excluded might actually be interesting.

Some links that I would not have expected:

-- Daniel Mietchen (talk) 03:30, 30 April 2020 (UTC)[reply]

Thanks for your feedback. The two items you mention also stuck out (while most of the others seemed to have some connection to covid). at the moment I am not sure what is the reason they receive such a high similarity score. I will also explore ways in which to share (some of) the data. Will keep you updated --MGerlach (WMF) (talk) 08:06, 30 April 2020 (UTC)[reply]

Did some digging and The Eyes of Darkness was part of a "did this book predict Covid-19" theory (snopes) and Typhoon Sarika has received very few page views so probably an anomaly. --Isaac (WMF) (talk) 19:17, 30 April 2020 (UTC)[reply]

Interesting. Thanks for digging up the story about the connection between covid and the book. Regarding the other one, checking the what links here for the typhoon sarika article shows up a bunch of local articles related to the coronavirus pandemic in Phillipines, Metro Manila, etc. Not sure if I understand correctly because I cant immediately find the link to typhoon sarika on those pages, but might suggest it is not an anomaly. --MGerlach (WMF) (talk) 18:35, 1 May 2020 (UTC)[reply]

Looks like the link is via the Rodrigo Duterte template at the end of the articles. Which suggests to me either that the sessions were bot-related but undetected (i.e. an indexing bot just gathering links and following them), an anomaly where a few people really did manage to discover and follow that link or happened to view both articles at a similar time, or some sort of second-order interaction that is being picked up by the model (e.g., Typhoon Sarika is like Philippines which is like 2020 coronavirus pandemic in the Philippines). The second-order interactions component of the model makes the most sense to me as an explanation, especially because the Typhoon page did not receive many pageviews so the model was using very little data to learn its representation. --Isaac (WMF) (talk) 21:02, 4 May 2020 (UTC)[reply]

Use in CDSC COVID Digital Observatory Project[edit]

Greetings!

I'm using this list as a source of keywords for searching for pandemic related discussions on social media as part of the Community Data Science Collective's COVID-19 Digital Observatory project.

Just thought you might like to know. You're welcome to drop in to chat at #communitydata-covid19@oftc.net. Groceryheist (talk) 21:52, 19 October 2020 (UTC)[reply]