Question

我对numpy非常陌生。

只是想知道为什么这不起作用。

print items['description']

产量

0                            Продам Камаз 6520 20 тонн
1                                      Весь в тюнинге.
2    Телефон в хорошем состоянии, трещин и сколов н...
3    Отличный подарок на новый год от "китайской ap...
4        Лыжные ботинки в хорошем состоянии, 34 размер
Name: description, dtype: object

尝试将此方法应用于此col中的所有行。

items['description'] = vectorize_sentence(items['description'].astype(str))

这是vectorize句子的函数定义。

def vectorize_sentence(self, sentence):
    # Tokenize 
    print 'sentence', sentence

    tkns = self._tokenize(sentence)
    vec = None
    for tkn in tkns: 
        print 'tkn', tkn.decode('utf-8')
        print type(tkn)
        if self.model[tkn.decode('utf-8')]:
            vec = sum(vec, self.model[tkn.decode('utf-8')])
    #vec = sum([self.model[x] for x in tkns if x in self.model])
    #print vec
def _tokenize(self, sentence):
    return sentence.split(' ')

错误讯息：

AttributeError: 'Series' object has no attribute 'split'

Answer 1

您收到该错误的原因是'Series' object has no attribute 'split'。主要是，.astype(str)不会像您认为的那样返回单个长字符串

items = pd.DataFrame({'description': ['bob loblaw', 'john wayne', 'lady gaga loves to sing']})
sentence = items['description'].astype(str)
sentence.split(' ')

现在尝试

sentence = ' '.join(x for x in items['description'])
sentence.split(' ')

然后在你的函数中实现

def _tokenize(self, sentence):
    return ' '.join(x for x in items['description']).split(' ')

将函数应用于numpy数组中的所有列

1 个答案: