我们应该如何使用 pad_sequences 在 keras 中填充文本序列?

How should we pad text sequence in keras using pad_sequences?

我使用从网络教程中获得的知识和我自己的直觉编写了一个 在 keras 中学习 LSTM 的代码。我将示例文本转换为序列,然后使用 keras 中的 pad_sequence 函数进行填充。

from keras.preprocessing.text import Tokenizer,base_filter
from keras.preprocessing.sequence import pad_sequences

def shift(seq, n):
    n = n % len(seq)
    return seq[n:] + seq[:n]

txt="abcdefghijklmn"*100

tk = Tokenizer(nb_words=2000, filters=base_filter(), lower=True, split=" ")
tk.fit_on_texts(txt)
x = tk.texts_to_sequences(txt)
#shifing to left
y = shift(x,1)

#padding sequence
max_len = 100
max_features=len(tk.word_counts)
X = pad_sequences(x, maxlen=max_len)
Y = pad_sequences(y, maxlen=max_len)

经过仔细检查,我发现我填充的序列看起来像这样

>>> X[0:6]
array([[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, 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, 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, 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, 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, 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, 0, 0, 0, 3],
       [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, 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, 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, 0, 0, 0, 2],
       [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, 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, 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, 0, 0, 0, 5],
       [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, 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, 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, 0, 0, 0, 4],
       [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, 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, 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, 0, 0, 0, 7]], dtype=int32)
>>> X
array([[ 0,  0,  0, ...,  0,  0,  1],
       [ 0,  0,  0, ...,  0,  0,  3],
       [ 0,  0,  0, ...,  0,  0,  2],
       ..., 
       [ 0,  0,  0, ...,  0,  0, 13],
       [ 0,  0,  0, ...,  0,  0, 12],
       [ 0,  0,  0, ...,  0,  0, 14]], dtype=int32)

填充后的序列应该是这样的吗?除数组中的最后一列外,其余均为零。我想我在按顺序填充文本时犯了一些错误,如果是的话,你能告诉我我在哪里犯了错误吗?

问题出在这一行:

tk = Tokenizer(nb_words=2000, filters=base_filter(), lower=True, split=" ")

当您设置这样的拆分(按 " ")时,由于您的数据的性质,您将得到由一个单词组成的每个序列。这就是为什么您的填充序列只有一个 non-zero 元素。要更改该尝试:

txt="a b c d e f g h i j k l m n "*100

如果你想按字符分词,你可以手动完成,不太复杂:

首先为你的角色建立一个词汇表:

txt="abcdefghijklmn"*100
vocab_char = {k: (v+1) for k, v in zip(set(txt), range(len(set(txt))))}
vocab_char['<PAD>'] = 0

这将为您的文本中的每个字符关联一个不同的数字。应保留索引为 0 的字符用于填充。

拥有反向词汇表将有助于解码输出。

rvocab = {v: k for k, v in vocab.items()}

有了这个之后,您可以先将文本拆分成序列,假设您想要长度为 seq_len = 13 的序列:

[[vocab_char[char] for char in txt[i:(i+seq_len)]] for i in range(0,len(txt),seq_len)]

您的输出将如下所示:

[[9, 12, 6, 10, 8, 7, 2, 1, 5, 13, 11, 4, 3], 
 [14, 9, 12, 6, 10, 8, 7, 2, 1, 5, 13, 11, 4],
 ...,
 [2, 1, 5, 13, 11, 4, 3, 14, 9, 12, 6, 10, 8], 
 [7, 2, 1, 5, 13, 11, 4, 3, 14]]

请注意,最后一个序列的长度不同,您可以将其丢弃或将您的序列填充为 max_len = 13,它会向其添加 0。

您可以用同样的方式构建目标 Y,将所有内容都移动 1。:-)

希望对您有所帮助。

参数 padding 控制每个序列之前或之后的填充。像这样使用:

X = pad_sequences(x, maxlen=max_len, padding='post')
Y = pad_sequences(y, maxlen=max_len, padding='post')