I'm training a Language Model using NLTK library of Python.
To obtain a better result, I use the Laplace smoothing technique.
But when I increase the N of N-gram model, my perplexity increases too, and I was expecting that the perplexity would descrease.
Is this behaviour normal? It happened because of te Laplace smoothing? Or am I making some mistake?
Example:
output perplexity using bigram: PP(an apple):3.7797631496846202 PP(an ant):5.738793548317168
output perplexity using trigram: PP(an apple):7.000000000000003 PP(an ant):7.000000000000003
bigram:
import nltk
from nltk.lm.preprocessing import padded_everygram_pipeline
from nltk.lm import MLE
from nltk.lm import Vocabulary
train_sentences = ['an apple', 'an orange']
tokenized_text = [list(map(str.lower, nltk.tokenize.word_tokenize(sent))) for sent in train_sentences]
n = 2
train_data = [nltk.bigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
words = [word for sent in tokenized_text for word in sent]
words.extend(["<s>", "</s>"])
padded_vocab = Vocabulary(words)
model = Laplace(n)
model.fit(train_data, padded_vocab)
test_sentences = ['an apple', 'an ant']
tokenized_text = [list(map(str.lower, nltk.tokenize.word_tokenize(sent))) for sent in test_sentences]
test_data = [nltk.bigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
for test in test_data:
print ("MLE Estimates:", [((ngram[-1], ngram[:-1]),model.score(ngram[-1], ngram[:-1])) for ngram in test])
test_data = [nltk.bigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
for i, test in enumerate(test_data):
print("PP({0}):{1}".format(test_sentences[i], model.perplexity(test)))
trigram:
n = 3
train_data = [nltk.bigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
words = [word for sent in tokenized_text for word in sent]
words.extend(["<s>", "</s>"])
padded_vocab = Vocabulary(words)
model = Laplace(n)
model.fit(train_data, padded_vocab)
test_sentences = ['an apple', 'an ant']
tokenized_text = [list(map(str.lower, nltk.tokenize.word_tokenize(sent))) for sent in test_sentences]
test_data = [nltk.trigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
for test in test_data:
print ("MLE Estimates:", [((ngram[-1], ngram[:-1]),model.score(ngram[-1], ngram[:-1])) for ngram in test])
test_data = [nltk.trigrams(t, pad_right=True, pad_left=True, left_pad_symbol="<s>", right_pad_symbol="</s>") for t in tokenized_text]
for i, test in enumerate(test_data):
print("PP({0}):{1}".format(test_sentences[i], model.perplexity(test)))