这是一个白盒答案：
使用您的原始代码，它会输出：
Traceback (most recent call last):
File "test.py", line 17, in <module>
unlabeled_featuresets)
File "/usr/local/lib/python2.7/dist-packages/nltk/classify/positivenaivebayes.py", line 108, in train
for fname, fval in featureset.items():
AttributeError: 'list' object has no attribute 'items'
查看第 17 行：
classifier = PositiveNaiveBayesClassifier.train(positive_featuresets,
unlabeled_featuresets)
似乎PositiveNaiveBayesClassifier 需要一个具有'.items()' 属性的对象，如果NLTK 代码是pythonic，那么直观上它应该是dict。
查看https://github.com/nltk/nltk/blob/develop/nltk/classify/positivenaivebayes.py#L88，对于positive_featuresets 参数应该包含的内容没有任何明确的解释：
:param positive_featuresets：被称为特征集的列表
正例（即他们的标签是True）。
检查文档字符串，我们看到了这个例子：
Example:
>>> from nltk.classify import PositiveNaiveBayesClassifier
Some sentences about sports:
>>> sports_sentences = [ 'The team dominated the game',
... 'They lost the ball',
... 'The game was intense',
... 'The goalkeeper catched the ball',
... 'The other team controlled the ball' ]
Mixed topics, including sports:
>>> various_sentences = [ 'The President did not comment',
... 'I lost the keys',
... 'The team won the game',
... 'Sara has two kids',
... 'The ball went off the court',
... 'They had the ball for the whole game',
... 'The show is over' ]
The features of a sentence are simply the words it contains:
>>> def features(sentence):
... words = sentence.lower().split()
... return dict(('contains(%s)' % w, True) for w in words)
We use the sports sentences as positive examples, the mixed ones ad unlabeled examples:
>>> positive_featuresets = list(map(features, sports_sentences))
>>> unlabeled_featuresets = list(map(features, various_sentences))
>>> classifier = PositiveNaiveBayesClassifier.train(positive_featuresets,
... unlabeled_featuresets)
现在我们找到了feature()函数，它将句子转换为特征并返回
dict(('contains(%s)' % w, True) for w in words)
基本上这是可以调用.items()的东西。看看字典理解，'contains(%s)' % w 似乎有点多余，除非它是为了人类可读性。所以你可以直接使用dict((w, True) for w in words)。
此外，用下划线替换空格也可能是多余的，除非以后会用到它。
最后，切片和有限迭代可以替换为可以提取字符 ngram 的 ngram 函数，例如
>>> word = 'alexgao'
>>> split=3
>>> [word[start:start+split] for start in range(0, len(word)-2)]
['ale', 'lex', 'exg', 'xga', 'gao']
# With ngrams
>>> from nltk.util import ngrams
>>> ["".join(ng) for ng in ngrams(word,3)]
['ale', 'lex', 'exg', 'xga', 'gao']
您的特征提取功能可以这样简化：
from nltk.util import ngrams
def three_split(word):
return dict(("".join(ng, True) for ng in ngrams(word.lower(),3))
[出]：
{'im ': True, 'm s': True, 'jim': True, 'ilv': True, ' si': True, 'lva': True, 'sil': True}
False
事实上，NLTK 分类器用途广泛，您可以使用字符元组作为特征，因此在提取特征时无需修补 ngram，即：
from nltk.classify import PositiveNaiveBayesClassifier
import re
from nltk.util import ngrams
chinese_names = ['gao', 'chan', 'chen', 'Tsai', 'liu', 'Lee']
nonchinese_names = ['silva', 'anderson', 'kidd', 'bryant', 'Jones', 'harris', 'davis']
def three_split(word):
return dict(((ng, True) for ng in ngrams(word.lower(),3))
positive_featuresets = list(map(three_split, chinese_names))
unlabeled_featuresets = list(map(three_split, nonchinese_names))
classifier = PositiveNaiveBayesClassifier.train(positive_featuresets,
unlabeled_featuresets)
print three_split("Jim Silva")
print classifier.classify(three_split("Jim Silva"))
[出]：
{('m', ' ', 's'): True, ('j', 'i', 'm'): True, ('s', 'i', 'l'): True, ('i', 'l', 'v'): True, (' ', 's', 'i'): True, ('l', 'v', 'a'): True, ('i', 'm', ' '): True}