TooAngel-Wave it based on self learning from conversations with a weighting function, which allows to select best fitting answers.

Currently it just can "replay" conversations. In every wave it participates it stores or updates the new incoming blips in its database, the more often they occur the higher rating. The relation between the blips is stored, too

If a new blip reaches tooangel-wave it searches the database for know relations and if it find some, it answer with it.

tooangel-wave should evolve to a conversation robot like Alice, but not so pushy.