You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current theory is that we can get away with 'good' suggestions with a simpler algorithm, though we might want something more complicated or differently simple.
At the moment we score people based on the number of rooms we share with them, inversely (smaller rooms == more likely to suggest those members) then alphabetically. We could track mentions, though that gets a bit complicated for a proof of concept. We'd probably want some generic scoring system for cross-client suggestions being stable, with a bonus for being mentioned "recently".
The text was updated successfully, but these errors were encountered:
Fixeselement-hq/element-web#11198
Note this doesn't implement the entire algorithm in 11198 because it feels too complicated at this stage. Instead, the idea is to review the suggestions closer to when the whole dialog is complete and fix them then: element-hq/element-web#11769
Algorithm for picking members is largely based on db5218e
yea, we need to do this before going live with the feature. The suggestions are terrible. I really don't want to start a chat with my dead-for-2-years IoT weather bot.
See #11198 (comment) for details.
The current theory is that we can get away with 'good' suggestions with a simpler algorithm, though we might want something more complicated or differently simple.
At the moment we score people based on the number of rooms we share with them, inversely (smaller rooms == more likely to suggest those members) then alphabetically. We could track mentions, though that gets a bit complicated for a proof of concept. We'd probably want some generic scoring system for cross-client suggestions being stable, with a bonus for being mentioned "recently".
The text was updated successfully, but these errors were encountered: