I use collected text from a channel on a Mumble server to train a Markov chain generator. My implementation isn't the best, but it is functional. I need more practice.
-
Install requirements with
$ pip install -r requirements.txt -
You'll need punkt and words bundles for NLTK I believe (correct me if I'm wrong). So do...
>>> import nltk >>> nltk.download()
-
Follow the instructions from there to download the appropriate modules.
-
Get a text file full of sample stuff to train it on and run it:
$ python markov.py stuff.txt
| Name | Description |
|---|---|
-h --help |
Show this screen. |
-l --length=<len> |
Set n-gram length [default: 2]. |
--limit=<lim> |
Total number of words max per sentence. |
--char=<lim> |
Maximum characters for generated sentence. |
-p --pickle |
Load training file as pickled data instead. |
-s --save <pickle> |
Save database as pickled data. |
-u --update <pickle> |
Update pickled database with given training file. |
-d |
Don't generate any sayings. |
<training_file> |
File for use in building the n-gram cache. |
[<n>] |
Number of sentences to generate. |