Smart collections: Automatically categorize pages into collections.

Categorize pages into collections using https://github.com/tensorflow/tfjs

Allow us to specify rules to automatically group existing pages/incoming pages based on certain keywords.

I’m less intrigued by a machine learning approach to this than a way to define filter rules for automatic tagging/categorization/automation, much like gmail has. A ML-backed dynamic rule creator could be built on top of that, but based on my experiences with google’s own attempts to decide automatically what is important/unimportant, I’d say it’s much easier to get good results by manually curating some rules.

4 Likes

I’d love to see this feature as well, though I’d consider it lower on my list of priorities since we do have full-text search. If we were only searching the titles/descriptions of pages I’d see this as being more essential.

If you know how to code you can already do that with StorexHub:
https://worldbrain.github.io/storex-docs/#/storex-hub/guides/memex/

There you could e.g listen to all pages that are tagged with “housing”, or that have those terms, and automatically add them to a specific collection.

If you run into troubles you can ping @v.denboer

I think the solution to this is the lowest-tech one: forget about the filters and machine learning and take advantage of your own natural logic and rhythms that are already being used when searching and opening pages. This is more or less how the chrome extension Tabs Outliner works, which I just posted about in this thread (Which other productivity apps are you using & which ones are you paying for (and why)?)

As you open tabs, relatively logical hierarchies/collections are automatically created (search for task managers or falafel recipes and all of the links you open from that search become children of that search. Any further links opened become sub-children). As you close tabs, they are discarded (do we really want to be saving, let alone cataloguing, 90+% of pages that we go to?). The key is to leave the useful pages open and then “archive” them by closing the window through the Tab Outliner tool, and then drag your “curated” hierarchies around a bit to tidy them up, rename things, etc… Its a nearly painless AND faultless system, something that filters, machine learning or manual organization cannot lay claim to.

Here’s an overview video of how it works: https://www.youtube.com/watch?v=OqjcrfKjobY