在 Python 中查找 tf-idf 时出现 fit_transform 错误

Question

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
mylist = [
    'a a b c',
    'a c c c d e f',
    'a c d d d',
    'a d f',
]
df = pd.DataFrame({"texts": mylist})
tfidf_vectorizer = TfidfVectorizer(ngram_range=[1, 1])
tfidf_separate = tfidf_vectorizer.fit_transform(df["texts"])

I am trying to find tf-idf value for “d” in line 3. But, it is showing me empty vocabulary error "ValueError: empty vocabulary; perhaps the documents only contain stop words".

Any advice on how to resolve the error would be appreciated!

Answer 1

You can do it like this:

• define analyzer='char' so that TfidfVectorizer works with the letters;
• find the index of d in the vocabulary and use it

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
mylist = [
    'a a b c',
    'a c c c d e f',
    'a c d d d',
    'a d f',
]
df = pd.DataFrame({"texts": mylist})
tfidf_vectorizer = TfidfVectorizer(ngram_range=[1, 1], analyzer='char')
tfidf_separate = tfidf_vectorizer.fit_transform(df["texts"])
ind = tfidf_vectorizer.vocabulary_['d']
tfidf_separate.todense()[2, ind]
>>> 0.6490674853546846