needing to ask: http://stackoverflow.com/a/37507447/821780

tweets.py

import nltk
from nltk.probability import LidstoneProbDist
from nltk.model.ngram import NgramModel 
import pandas as pd

tweets = pd.read_csv('tweeets.csv')
tokenize = lambda x: nltk.word_tokenize(str(x))
train = [tokenize(text) for text in tweets[tweets.user == 'trylks']['text']]
text = tokenize("I think that the #Python library #nltk is great")

# here is the good part
estimator = lambda fdist, bins: LidstoneProbDist(fdist, 0.2)
model = NgramModel(3, train, estimator=estimator)
perplexity = model.perplexity(text)