2

I made 2 and 3-gram models from my text file.

from nltk import *
text = open('Alice in Wonderland.txt', 'r').read()
table = string.maketrans('', '')
text = text.translate(table, string.punctuation)
tokens = word_tokenize(text.lower())
bigram = nltk.bigrams(tokens)
trigram = nltk.trigrams(tokens)

but how can I generate new sentences using these models?

alvas
  • 115,346
  • 109
  • 446
  • 738
Mohammad Hassan Sohan Ajini
  • 150
  • 1
  • 12

1 Answers1

2

Currently, NLTK's generate() function is being deprecated because it is broken, see https://github.com/nltk/nltk/issues/1180

But a state-of-art alternative is text generation using Recurrent Neural Nets, e.g. https://github.com/karpathy/char-rnn (Note: Unlike traditional Ngram based hidden markov model, the char-RNN doesn't use ngrams information.)

Alternatively you can implement your own hidden markov model, see http://fulmicoton.com/posts/shannon-markov/

alvas
  • 115,346
  • 109
  • 446
  • 738