Introduction to Feedseer

Feedseer is a Mastodon instance that gives users the ability to automatically filter out/categorize toots based on their content. It can be used to do much more than regex based filtering. Filtering is not enabled by default.

How is Feedseer different from Mastodon?

Feedseer’s qualifier store gives you the superpowers needed to:

  • Ignore abusive comments from users you don’t follow. Check out the Abuse qualifier in our store.
  • Filter out toots on topics you are not interested in. Qualifier store has qualifiers on several topics.
  • Create qualifiers to suit your specific needs.

Qualifier store is similar to an App Store on your phone. Browse the store to install the qualifiers you like, review them or create your own qualifiers.

How does Feedseer let you control your timeline?

Qualifiers analyze the content of a toot and take action based on that. Based on a qualifier’s output, actions can be configured to either hide a toot or move it to a folder. Users can create their own qualifiers to analyze different types of toots or provide interesting insights on a toot’s content.

Feedseer’s qualifiers can do much more than regex filtering, like use AI to analyze the sentiment of a toot. The implementation details of a qualifier (regex, AI or other) are up to its creator and happen outside of Feedseer. Independently, a Feedseer user has the freedom to not use a qualifier (the default), or use it to take action (categorize/filter).

Here are a couple of examples.

  1. Ignore posts on topics you’re not interested in.

Let’s say, you are interested in the JavaScript programming language. To get updates on the language, you follow an amazing JavaScript developer called JavaScript guru (jsguru). The JavaScript guru also posts on European soccer. But you hate sports for some reason and want to filter out toots on sports. Here is how you can do it on Feedseer.

  • JavaScript guru tooting about soccer.
  • The soccer toot shows up in your timeline.
  • Let’s say that you want to move all sports related toots to a folder called Sports. You can create a folder from this menu.
  • Click the ‘+’ icon to create a folder.
  • Now, go to Qualifier Store.
  • Click on the Sports qualifier.
  • Install the qualifier.
  • Click on Add Filter button. Set Action type to Move to folder and select the Sports folder you created.
  • Click on Save Changes. The saved qualifier should look like this.
  • JavaScript guru toots on soccer again.
  • This toot will show up in your Sports folder, but not in your timeline.
  • Let’s say, you don’t want sport-related toots in your timeline and folders. Go to your Installed qualifiers.
You can uninstall a qualifier by clicking on the ‘x’ icon.
  • Change Action type in the Sports qualifier configuration to Skip Inbox.
  • Save your changes.
  • The next time JavaScript guru (or any of your friends) post on soccer…
  • That toot will not show up in your timeline or folders.
Note the timestamp of toots in your timeline and folders.
  • This works on toots with URLs also.

Note that the user could have installed Soccer qualifier if they wanted to filter out only soccer-related toots, but not all sport-related ones.

The current version of topic-based (sports, politics, etc.) qualifiers use IBM natural language APIs to find categories for text/URLs in a toot. It would be slightly harder to achieve the same result with filters since the user would have to create a filter with a lot of words on each topic. This is the USP of Feedseer — the ability to use an external service to construct your timeline.

2. Ignore angry toots

  • Let’s say there’s a crazy user who you don’t follow, but sends you messages.
  • Create a folder called Abuse.
  • Go to Qualifier store and install the Abuse qualifier.
  • Add a filter to move angry toots to a folder called Abuse.
  • Click on Save Changes.
  • The next angry toot by Crazy Person (or others) will go to a folder and not show up in your notifications. Note the most recent toot (at the top of timeline on right).
Crazy person’s toots
The most recent toot was moved to a folder and the user was not notified.

Abuse qualifier uses the IBM Tone Analyzer API to find anger in toots.

Conclusion and future work

The two examples might seem reductive, but I just wanted to give a high-level overview of what qualifiers are currently capable of. Users can create their own qualifiers to identify other topics or emotions.

Of course, there is room for a lot of improvement like applying qualifiers selectively to users/lists, scaling, privacy and many others. Please feel free to suggest ideas or contribute.

Source: https://github.com/sunilbandla/mastodon

Thanks a lot!