Hermes Module
Create a module, similar to markov in Supybot, which has the ability to simulate natural speech. This module should also be able to provide useful information and language variability instead of the random spews of random stuff which markov often spits out.
Blueprint information
Related branches
Related bugs
Sprints
Whiteboard
I'm in support for this, assuming that we first make an interactive makeconf script, so we can include the module in the user config.cfg file.
Having said that, how would you, theoretically, attempt to match the variability and useful information provider to the intricacies of human speech? Because the limitation with markov bots is that they sometimes fail.
--EvilPhoenix
-------
I was planning on going with what you could call overly complicated markov chains. I think that the ones the markov module uses are far too simple. For example, the bot could learn various ways to say hello. It would then choose at random between all the ways it knows.
It will also have a threshold of responses required before something becomes valid. One person saying something after someone else says "lol" doesn't make it a valid response to "lol."
Lastly, I think I could code a method to have the bot "follow" conversations. If two people are highlighting each other while they are talking, we can guess that they are talking to each other and not the whole channel. So, this would also narrow down the number of out-of-place responses.
--AJ00200
-------
That's actually a very good idea. In theory, it would work, but in practice, it would need some kind of training period for it, no?
--EvilPhoenix
-------
Yes, definitely. However, I hope to counter this by making database files available for download. That way, they can start their bots off with the combined intelligence of the BBot Core Developers.
--AJ00200
