aperiodic / clj-markov Goto Github PK

Create a generate-sentence function that randomly chooses from among the states of the chain that end in a period, and generate tokens until another period is reached or some ceiling amount of tokens has been generated.

Use OpenNLP to tokenize corpuses for training and de-tokenize generated tokens for display.

Create a train-corpuses function that accepts multiple filenames, tokenizes them, calculates normalized weights based on corpus size, and returns a Markov chain trained on the corpuses with the normalized weights.

Tag tokens by their source corpus in order to be able to bias towards generating sentences that cross from one corpus to another.