With my slightly OCD tendency, I often trawl through topics adding tags.
Opening the tag drop-down shows five popular tags (I assume, the most popular?).
Here’s an idea - how about making the suggested tags a bit smarter using some basic machine learning? I’d suggest the following:
Background Job
Discourse runs a background job that analyses the text and tags of recent topics using a Naive Bayes Classifier
-
The background job uses existing tags on topics to build a model. It learns which tags are likely to be relevant to text content.
-
For topics without tags, the background job runs the text through the built model to find any tags with a high probability of applying to the topic. These are persisted in the DB as “suggested tags” for that topic.
Suggested Tag UX
When a user edits the title/tags, the UI shows any strong tag suggestions, and the user can click on a suggested tag to add it to the topic