如何在python或R中获得最常见的短语或单词

Question

Given some text, how can i get the most common n-gram across n=1 to 6? I've seen methods to get it for 3-gram, or 2-gram, one n at a time, but is there any way to extract the max-length phrase that makes the most sense, and all the rest too?

for example, in this text for demo-purpose only: fri evening commute can be long. some people avoid fri evening commute by choosing off-peak hours. there are much less traffic during off-peak. fri evening commute can be long. some people avoid fri evening commute by choosing off-peak hours. there are much less traffic during off-peak.

The ideal outcome of n-gram and their counter would be:

fri evening commute: 3,
off-peak: 2,
rest of the words: 1

any advice appreciated. Thanks.

Answer 1

Python

Consider the NLTK library which offers an ngrams function that you can use to iterate over values of n.

A rough implementation would be along the lines of the following, where rough is the keywords here:

from nltk import ngrams
from collections import Counter

result = []
sentence = 'fri evening commute can be long. some people avoid fri evening commute by choosing off-peak hours. there are much less traffic during off-peak.'
# Since you are not considering periods and treats words with - as phrases
sentence = sentence.replace('.', '').replace('-', ' ')

for n in range(len(sentence.split(' ')), 1, -1):
    phrases = []

    for token in ngrams(sentence.split(), n):
        phrases.append(' '.join(token))

    phrase, freq = Counter(phrases).most_common(1)[0]
    if freq > 1:
        result.append((phrase, n))
        sentence = sentence.replace(phrase, '')

for phrase, freq in result:
    print('%s: %d' % (phrase, freq))

As for R

This might be helpful

Answer 2

如果您打算使用R，我建议您这样做： https : //cran.r-project.org/web/packages/udpipe/vignettes/udpipe-usecase-postagging-lemmatisation.html