Open joeclark-phd opened 5 years ago
One cool thing I saw in his demo site was that he would blend two input files in training a Markov generator, so you could get text based on (for example) national capitals blended with Pokemon characters. I might refactor the train()
method to allow continuing training with multiple input streams so we could easily do this, too. (Currently that method erases its model and restarts training when called.)
I found a similar project written in Haxe by github user Tw1ddle: https://github.com/Tw1ddle/MarkovNameGenerator. He's got some great examples including a demo site, which might be valuable for our Java implementation. Take a look at it and see if we can use it.