如何获取 pandas 数据框中单词列表（子字符串）的出现次数？

Question

我有一个 pandas 数据框，大约有 150 万行。我想在某一列中找到特定的、选定的单词（都是已知的）的出现次数。这适用于单个单词。

d = df["Content"].str.contains("word").value_counts()

但我想从列表中找出多个已知词的出现次数，例如 "word1"、"word2"。 word2 也可以是 word2 或 wordtwo，像这样：

word1           40
word2/wordtwo   120

我该如何完成？

Answer 1

IMO 最有效的方法之一是使用 sklearn.feature_extraction.text.CountVectorizer 向其传递一个词汇表（单词列表，您想要计数）。

演示：

In [21]: text = """
    ...: I have a pandas data frame with approximately 1.5 million rows. I want to find the number of occurrences of specific, selected words in a certain colu
    ...: mn. This works for a single word. But I want to find out the occurrences of multiple, known words like "word1", "word2" from a list. Also word2 could
    ...: be word2 or wordtwo, like so"""

In [22]: df = pd.DataFrame(text.split('. '), columns=['Content'])

In [23]: df
Out[23]:
                                             Content
0  \nI have a pandas data frame with approximatel...
1  I want to find the number of occurrences of sp...
2                       This works for a single word
3  But I want to find out the occurrences of mult...
4      Also word2 could be word2 or wordtwo, like so

In [24]: from sklearn.feature_extraction.text import CountVectorizer

In [25]: vocab = ['word', 'words', 'word1', 'word2', 'wordtwo']

In [26]: vect = CountVectorizer(vocabulary=vocab)

In [27]: res = pd.Series(np.ravel((vect.fit_transform(df['Content']).sum(axis=0))),
                         index=vect.get_feature_names())

In [28]: res
Out[28]:
word       1
words      2
word1      1
word2      3
wordtwo    1
dtype: int64

Answer 2

您可以像这样创建字典：

{w: df["Content"].str.contains(w).sum() for w in words}

假设 words 是单词列表。