I've made the difficult decision to turn off the US Politics Labeler. The linked post has more info about why and what's next. I will shortly begin the process of removing the labeler's data from my server. If you want to run your own, the source code is linked in the article.

As a reminder, appeals for posts with the "no alt text" label are not accepted and will be dismissed immediately. It simply checks for the presence of alt text, and you can easily avoid it by adding appropriate alt text to your images. This label will remain until a better solution can be found.

Due to a misconfiguration, the automatic labeling script may not have been running the past few days (or running in a very limited capacity). It may not be possible to fix until tomorrow due to rate limits. It is running in a limited capacity now. I'm very sorry for the inconvenience.

I somehow just found out that posts without alt text weren't getting labeled. They should start getting labeled now. Sorry about that.

I've just introduced a new label that allows you to block posts that have missing or bad alt text for their media. It defaults to "Warn" because I think people would get pissed that their timelines are suddenly and unexpectedly devoid of images. I recommend you set it to "Hide".

The service may see some downtime and more missed posts than usual as I am about to scale up the server. Donations to help cover the cost would be appreciated. liberapay.com/bleonard252

The source code for this service's automation tool, codenamed Star Spangled, has been released on Codeberg. This allows you to set up your own powerful weighted-keyword labeler. A full guide on its installation along with Ozone will be written in the future. codeberg.org/bleonard252/...

Idea: what if I introduced a label, ignored by default, that applies to posts that have media without alt text?

I just installed another update, which will also crawl the reply-threads of posts it finds, up and down. If that feature fails, it should not affect any other operations of the labeler.

Just pushed an update that should mean less missed reports. There may be hiccups due to the server restart; I'll keep you posted.

As a reminder, we don't label full accounts unless they belong to a politician or political organization. Check the policy document for the full list of exceptions.

We don't label troll accounts. You should instead report them to a blocklist maintainer.

Should I start labeling MAGA trolls? 1️⃣ Yes 2️⃣ No 📊 Show results

If you notice an incorrectly labeled post (wrong label? shouldn't be labeled?) you can appeal or report it, which puts it to manual review. This helps me figure out ways to improve automatic labeling to correct the issue in the future.

The auto-labeler is now coming back online, and reports are being addressed. We sincerely apologize for the downtime, and measures have been taken and tested successfully to make sure that this does not happen again.

The auto-labeler is now coming back online, and reports are being addressed. We sincerely apologize for the downtime, and measures have been taken and tested successfully to make sure that this does not happen again.

Heads-up: the auto-labeler will be turned off overnight due to an issue. Measures have been taken to prevent this issue in the future, but the downtime today is unfortunately unavoidable. We're deeply sorry for the inconvenience this will cause.

I've just pushed an update to the script which will more accurately handle phrases (multi-word "keywords"). This should mean fewer false positives. If you still believe I've labeled your post in error, appeal it, and I should be able to diagnose why it was labeled, and fix it.

After auditing a profile, I decided it was sufficient and necessary to label the entire account. The account in question posts exclusively political images, which are not yet possible to automatically label.

An update was pushed to listen to more sources for labeling. This should mean that more posts are already labeled appropriately before you even see them. As always, report anything we missed and appeal false positives. 🖖

For some reason the automatic labeling script cannot log into the account, and therefore cannot check for reports or label anything. Diagnosing.

Some of the labels (particularly the LGBTQIA+ label) don't have keywords set up yet. I might come back in and do those tomorrow. The labeler automatically tries to handle every report. The report stays up if it doesn't match, so it can be manually handled.

I know it's too little too late but I channeled my election anxiety to finally set up the automator for @uspol.blakeslabs.com! Send in reports for anything you feel needs the label.

Here is a link to our current (original) labeling policy, an early draft providing guidance and transparency on our decisions.

I've set the "Election" label to "alert" severity, which I believe means you get an even angrier warning than usual on election related posts. I expect to revert this specific change later this month.

Right now there's literally nothing labeled with this labeler yet. I think sometime later I'll hook up a script that looks for keywords, and reports those. I think I'll also do it with score and maybe some anti-filter-evasion.