NLTK标记化“不可损坏类型：'list'”

Question

以下示例：Twitter data mining with Python and Gephi: Case synthetic biology

CSV to: df['Country', 'Responses']

'Country'
Italy
Italy
France
Germany

'Responses' 
"Loren ipsum..."
"Loren ipsum..."
"Loren ipsum..."
"Loren ipsum..."

1. 标记“响应”中的文本
2. 删除100个最常见的单词（基于brown.corpus）
3. 找出剩下的100个最常用的单词

我可以完成步骤1和2，但在步骤3中出现错误：

TypeError: unhashable type: 'list'

我相信这是因为我在一个数据帧中工作，并进行了以下（可能是有害的）修改：

原始示例：

#divide to words
tokenizer = RegexpTokenizer(r'\w+')
words = tokenizer.tokenize(tweets)

我的代码：

#divide to words
tokenizer = RegexpTokenizer(r'\w+')
df['tokenized_sents'] = df['Responses'].apply(nltk.word_tokenize)

我的完整代码：

df = pd.read_csv('CountryResponses.csv', encoding='utf-8', skiprows=0, error_bad_lines=False)

tokenizer = RegexpTokenizer(r'\w+')
df['tokenized_sents'] = df['Responses'].apply(nltk.word_tokenize)

words =  df['tokenized_sents']

#remove 100 most common words based on Brown corpus
fdist = FreqDist(brown.words())
mostcommon = fdist.most_common(100)
mclist = []
for i in range(len(mostcommon)):
    mclist.append(mostcommon[i][0])
words = [w for w in words if w not in mclist]

Out: ['the',
 ',',
 '.',
 'of',
 'and',
...]

#keep only most common words
fdist = FreqDist(words)
mostcommon = fdist.most_common(100)
mclist = []
for i in range(len(mostcommon)):
    mclist.append(mostcommon[i][0])
words = [w for w in words if w not in mclist]

TypeError: unhashable type: 'list'

有很多问题都列在不好的清单上，但据我所知，没有一个问题是完全相同的。 有什么建议吗？谢谢

---

回溯

TypeError                                 Traceback (most recent call last)
<ipython-input-164-a0d17b850b10> in <module>()
  1 #keep only most common words
----> 2 fdist = FreqDist(words)
  3 mostcommon = fdist.most_common(100)
  4 mclist = []
  5 for i in range(len(mostcommon)):

/home/*******/anaconda3/envs/*******/lib/python3.5/site-packages/nltk/probability.py in __init__(self, samples)
    104         :type samples: Sequence
    105         """
--> 106         Counter.__init__(self, samples)
    107 
    108     def N(self):

/home/******/anaconda3/envs/******/lib/python3.5/collections/__init__.py in __init__(*args, **kwds)
    521             raise TypeError('expected at most 1 arguments, got %d' % len(args))
    522         super(Counter, self).__init__()
--> 523         self.update(*args, **kwds)
    524 
    525     def __missing__(self, key):

/home/******/anaconda3/envs/******/lib/python3.5/collections/__init__.py in update(*args, **kwds)
    608                     super(Counter, self).update(iterable) # fast path when counter is empty
    609             else:
--> 610                 _count_elements(self, iterable)
    611         if kwds:
    612             self.update(kwds)

TypeError: unhashable type: 'list'

Answer 1

FreqDist函数接受可散列对象的iterable（使其成为字符串，但它可能与任何对象一起工作）。您得到的错误是因为您传入了一个列表的iterable。正如您所建议的，这是因为您所做的更改：

df['tokenized_sents'] = df['Responses'].apply(nltk.word_tokenize)

如果我正确理解了Pandas apply function documentation，那么这一行将nltk.word_tokenize函数应用于某个系列word-tokenize返回单词列表

作为解决方案，在尝试应用FreqDist之前，只需将列表添加到一起，如下所示：

allWords = []
for wordList in words:
    allWords += wordList
FreqDist(allWords)

一个更完整的修订，做你想做的。如果您只需要识别第二组100，请注意mclist将在第二次识别

df = pd.read_csv('CountryResponses.csv', encoding='utf-8', skiprows=0, error_bad_lines=False)

tokenizer = RegexpTokenizer(r'\w+')
df['tokenized_sents'] = df['Responses'].apply(nltk.word_tokenize)

lists =  df['tokenized_sents']
words = []
for wordList in lists:
    words += wordList

#remove 100 most common words based on Brown corpus
fdist = FreqDist(brown.words())
mostcommon = fdist.most_common(100)
mclist = []
for i in range(len(mostcommon)):
    mclist.append(mostcommon[i][0])
words = [w for w in words if w not in mclist]

Out: ['the',
 ',',
 '.',
 'of',
 'and',
...]

#keep only most common words
fdist = FreqDist(words)
mostcommon = fdist.most_common(100)
mclist = []
for i in range(len(mostcommon)):
    mclist.append(mostcommon[i][0])
# mclist contains second-most common set of 100 words
words = [w for w in words if w in mclist]
# this will keep ALL occurrences of the words in mclist