如何在字符级别对句子进行单热编码？

Question

我想将一个句子转换为一个单热向量数组。这些向量将是字母表的单热表示。它看起来像下面这样：

"hello" # h=7, e=4 l=11 o=14

会变成

[[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
 [0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]

不幸的是，来自 sklearn 的 OneHotEncoder 不作为输入字符串。

Answer 1

您询问了 "sentences"，但您的示例只提供了一个词，所以我不确定您想对空格做什么。但就单个单词而言，您的示例可以通过以下方式实现：

def onehot(ltr):
 return [1 if i==ord(ltr) else 0 for i in range(97,123)]

def onehotvec(s):
 return [onehot(c) for c in list(s.lower())]

onehotvec("hello")
[[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]

Answer 2

只需将传递的字符串中的字母与给定的字母表进行比较即可：

def string_vectorizer(strng, alphabet=string.ascii_lowercase):
    vector = [[0 if char != letter else 1 for char in alphabet] 
                  for letter in strng]
    return vector

请注意，使用自定义字母表（例如 "defbcazk"，列将按照每个元素出现在原始列表中的顺序排列）。

string_vectorizer('hello')的输出：

[[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
 [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]

Answer 3

使用 pandas，您可以通过传递分类系列来使用 pd.get_dummies：

import pandas as pd
import string
low = string.ascii_lowercase

pd.get_dummies(pd.Series(list(s)).astype('category', categories=list(low)))
Out: 
   a  b  c  d  e  f  g  h  i  j ...  q  r  s  t  u  v  w  x  y  z
0  0  0  0  0  0  0  0  1  0  0 ...  0  0  0  0  0  0  0  0  0  0
1  0  0  0  0  1  0  0  0  0  0 ...  0  0  0  0  0  0  0  0  0  0
2  0  0  0  0  0  0  0  0  0  0 ...  0  0  0  0  0  0  0  0  0  0
3  0  0  0  0  0  0  0  0  0  0 ...  0  0  0  0  0  0  0  0  0  0
4  0  0  0  0  0  0  0  0  0  0 ...  0  0  0  0  0  0  0  0  0  0

[5 rows x 26 columns]

Answer 4

这是一个矢量化方法，使用 NumPy broadcasting 给我们一个 (N,26) 形状的数组 -

ints = np.fromstring("hello",dtype=np.uint8)-97
out = (ints[:,None] == np.arange(26)).astype(int)

如果您正在寻找性能，我建议使用初始化数组然后分配 -

out = np.zeros((len(ints),26),dtype=int)
out[np.arange(len(ints)), ints] = 1

样本运行-

In [153]: ints = np.fromstring("hello",dtype=np.uint8)-97

In [154]: ints
Out[154]: array([ 7,  4, 11, 11, 14], dtype=uint8)

In [155]: out = (ints[:,None] == np.arange(26)).astype(int)

In [156]: print out
[[0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0]]

Answer 5

这是循环神经网络中的一项常见任务，如果您想使用它，tensorflow 中有专门用于此目的的特定函数。

alphabets = {'a' : 0, 'b': 1, 'c':2, 'd':3, 'e':4, 'f':5, 'g':6, 'h':7, 'i':8, 'j':9, 'k':10, 'l':11, 'm':12, 'n':13, 'o':14}

idxs = [alphabets[ch] for ch in 'hello']
print(idxs)
# [7, 4, 11, 11, 14]

# @divakar's approach
idxs = np.fromstring("hello",dtype=np.uint8)-97

# or for more clear understanding, use:
idxs = np.fromstring('hello', dtype=np.uint8) - ord('a')

one_hot = tf.one_hot(idxs, 26, dtype=tf.uint8)
sess = tf.InteractiveSession()

In [15]: one_hot.eval()
Out[15]: 
array([[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=uint8)

如何在字符级别对句子进行单热编码？

How to one-hot-encode sentences at the character level?

python

nlp

numpy

pandas

one-hot-encoding