*A character level recurrent neural net is trained to generate tweets in the style of a set of existing tweets. *
Source tweets are assumed to be in a column named 'text' in "source_tweets.csv" and can be downloaded from twitter using a tool such as tweet_dumper.
To Train:
>>> network.fit(dummy_data, dummy_data)
To generate tweets:
>>> generate_tweets(network, count)
Running from the command line will attempt to load existing network data from "network.params" and continue training. When finished training, or interupted, the current network data is dumped to "network.params_out". If you want to continue training with this data next time, this file should be moved to "network.params".
Running dump_tweets(network, count, path) will dump count tweets to the locations
specified by path.