A discussion is ongoing about suggesting alternative categories based on words used in a post on Discourse. Sujan initiated the conversation, referencing a GitHub feature that automatically adds labels to issues based on their content. jomaxro mentioned that Discourse already checks post content for similarity, but does not suggest category changes. Sujan proposed defining a list of “trigger words” connected to specific categories, which could be implemented through a plugin. codinghorror suggested using Admin, Logs, Watched Words to take action on certain words. sam mentioned the possibility of leveraging AI models like ChatGPT to guess categories accurately.
You used the string “widget” in your post, but are trying to post in the “foo” category. Do you maybe want to change the category to “widget” so it matches the content of your post?
Does something like this (or similar) exist?
On Github for Visual Studio Code has this neat bot that adds labels based on what you wrote. So if you write about the “minimap” feature, it tags the issue with “minimap” automatically.
Discourse already checks the post content against posts to check for similarity (to avoid duplicate topics), but does not explicitly suggest a category change.
How do you propose this would work? From a technical perspective, what should Discourse compare the content of the post against to suggest a different category?
I was thinking of being able to define a list of “trigger words” for example that then are connected to a category that should be suggested:
(This is probably a task for a plugin, not core Discourse integration I think)
Well, you can set up Admin, Logs, Watched Words to take action on certain words as needed.
This is an interesting one these days, it is possible we could lean on chatgpt or similar to guess categories reasonably accurately