Now we do definely know that those are user name, we can probably add code specifically to handle these cases.
Right now the parser has no way to identify this information. It seems worth it to create a "rich message" format that ie slack server can encode the text versus the user name metadata as needed and then karma will be able to take advantage of this information.
For pre-existing services such as irc ,they will keep on passing in the text data without additional processing for richness.
This will allow karma to take the space in factor after the name for eg.