It's really easy to read at a glance. I like it.
Apr 19 2019
Apr 18 2019
Looks like joblib uses cloudpickle, so we should probably use that directly.
Looks good to me.
Apr 17 2019
Thank you very much @eranroz! Im excited to sharpen up the predictions for hewiki :D
I see the new version now.
"Prediction" represents the quality class with the highest probability (in this case, "GA" at 42%).
The number that you see represents the central mass of the probability distribution. In most cases these agree, but in some cases, there's a high level of uncertainty and that's what we're seeing here.
When I score that revision, I get:
Apr 16 2019
I think we're running into the same situation that we did on Basque wiki. Because the model has a high level of uncertainty, the "weighted sum" values are pulled towards the middle (3/4 in this case). This is just an artifact of the distribution of prediction probabilities. We can help ORES be more confidence by giving it more training data. I've been looking into this problem for euwiki, so maybe we could do a new labeling campaign in Wiki labels for glwiki too. What do you think? I'd be looking to have you (and other glwiki-pedians) label 500-2000 new pages to give ORES more examples.
Woops. EC, but it looks like we're working from similar revisions.
Aha. It looks like this version gets B (4.26).
I see. Can you tell me what version of the second article got the weird values?
Apr 14 2019
I see "ORES Valoración: AB (4.16)" for the first and "ORES Valoración: AC (5.23)" for the second. That seems right to me. Maybe there were changes made since you asked (sorry I've been away from keyboard for a few days).
I'm inspired by how the Teahouse uses "Guest" and "Host" instead of "Newcomer" and "Mentor". I'm not sure if "Guest" fits our context perfectly. Maybe "Seeker" as in a "Help seeker" or a "Project seeker".
Hey folks. I'm sorry to not respond sooner. I've been AFK for a tech management summit and I'm on vacation until Wednesday, April 17th. I've shared the document with everyone in this task with the hopes that changes could be made by others as well. Please feel free to make edits or I'll pick this up again when I get back on the 17th.
Apr 10 2019
Seems like this is done.
Apr 9 2019
Sorry this one as left hanging for a while. We'll be giving this priority and doing some exploration ASAP.
I juts had a look and we averaged 5.2k uncached scores per minute during our last overload event.
Apr 8 2019
@srodlund, this is now ready for your review. Once we have that, we can make changes to ores.wikimedia.org.
+1 LGTM. Please proceed.
BTW, I've been discussing this model on our talk page: Topic: What changes in probabilities are significant?
Apr 6 2019
Apr 5 2019
I just finished reviewing and making some changes. @srodlund, could you give us a final review? Then we'll update the actual homepage.
Apr 4 2019
Merged, but I don't see a new campaign yet. https://labels.wmflabs.org/ui/ruwiki/
Apr 3 2019
^ Changes our config to use the new DNS alias.
The admin actions were one of the primary reason for creating this Epic. I'm OK with debating the future, but it doesn't make sense to rewrite history.
Right. I don't want to close that because I think there's more for us to try (that I might pitch at the Hackathon. But I think we should call this general task done.
There is a sub-task that is not done. The creation of admin actions. Wikilabels is not "fully-featured". I think we need to decline this task if we're not going to be doing that last bit.
Apr 2 2019
@Catrope is this still a thing?
@Catrope is this still a thing?
Declining in favor Jade
Declining because we're following the rest of the org to K8s. We'll be considering streaming architecture more later. K8s and streaming are orthogonal.