AG_NEWS分类文本

AG_NEWS简介

简单描述：AG_NEWS是通过从原始语料库中选择4个最大的类别来构建的。每个类别包含30,000个训练样本和1,900个测试样本。训练样本的总数为120,000，测试样本为7,600。 https://github.com/mhjabreel/CharCNN/tree/master/data

简单文本处理介绍

from torchtext.vocab import build_vocab_from_iterator

# 假设这是一个包含文本分词结果的列表
word_list = ["apple", "banana", "crange", "dear", "grape","kiwi", "strawberry","<unk>", "watermelon"]

# 构建词汇表对象
vocab = build_vocab_from_iterator(iter([word_list]))

# 设置默认索引为 "<unk>" 对应的索引,默认索引也就是起始索引，下标为0
vocab.set_default_index(vocab["<unk>"])

print(vocab(['apple', 'banana']))

# 将一个包含未知单词的列表转换为索引列表，也就是默认索引为初始下标0，其余字母按照首字母大小排列
index_list = [vocab[token] for token in word_list]

# 打印转换后的索引列表
print(index_list)

vocab_size = len(vocab)#计算词汇表大小
print(vocab_size)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

输出结果

AG_NEWS文本分类开始啦

1.引入相关模块和文本预处理

from torchtext.datasets import AG_NEWS
import torch

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

train_iter = AG_NEWS(split='train')

from torchtext.data.utils import get_tokenizer
from torchtext.vocab import build_vocab_from_iterator

tokenizer  = get_tokenizer('basic_english') # 返回分词器函数，“get_tokenizer函数详解”在下述

#通过for循环遍历data_iter中的每个元素（即每个包含文本内容的元组）。由于只关心文本内容，因此使用下划线 _ 来表示元组中的第一个元素（通常是ID或索引），并将文本内容赋值给变量text.
# 然后，调用之前定义的分词器函数 tokenizer 对文本进行分词，并使用 yield 关键字将分词结果依次返回给调用者。
def yield_tokens(data_iter):
    for _, text in data_iter:#获取每一条的标签label和内容text
        yield tokenizer(text)#对获取内容分词，并返回。yield返回一个迭代器对象

for i in range(3):
    print(list(next(yield_tokens(train_iter))))

vocab = build_vocab_from_iterator(yield_tokens(train_iter), specials=["<unk>"])
vocab.set_default_index(vocab["<unk>"]) # 设置默认索引，如果找不到单词，则会选择默认索引

print(vocab(['here', 'is', 'an', 'example'])) #输出的是每个单词的索引

text_pipeline  = lambda x: vocab(tokenizer(x)) #接受参数x，返回的是vocab(tokenizer(x))分词索引
label_pipeline = lambda x: int(x) - 1 #接受参数x，返回该数值减一

print(text_pipeline('here is the an example'))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

PS：get_tokenizer() 函数通常是在某个 NLP（自然语言处理）库中定义的一个函数，它用于获取适当的分词器。分词是自然语言处理中的一个重要环节，它将一段文本切割成若干个有意义的单元，这些单元被称为“token”，是后续自然语言处理任务的基础。
不同的 NLP 库会提供不同的分词器，get_tokenizer() 函数就是用来获取相应的分词器。具体而言，这个函数会根据传入的参数（一般是语言或者语言加上特性），返回一个对应的分词器对象。

例如，假设我们使用了 PyTorch 的 torchtext 库，我们可以按照以下方式获取适当的英语分词器：
from torchtext.legacy.data.utils import get_tokenizer
tokenizer = get_tokenizer(‘basic_english’)
tokens = tokenizer(“Hello world, I’m an AI assistant.”)
print(tokens)
输出为
[‘hello’, ‘world’, ‘,’, ‘i’, “'m”, ‘an’, ‘ai’, ‘assistant’, ‘.’]

2.定义数据加载器

from torch.utils.data import DataLoader

def collate_batch(batch):
    label_list, text_list, offsets = [], [], [0]

    for (_label, _text) in batch:
        # 标签列表
        label_list.append(label_pipeline(_label))

        # 文本列表
        processed_text = torch.tensor(text_pipeline(_text), dtype=torch.int64)
        text_list.append(processed_text)

        # 偏移量，即语句的总词汇量
        offsets.append(processed_text.size(0))

    label_list = torch.tensor(label_list, dtype=torch.int64)
    text_list = torch.cat(text_list)
    offsets = torch.tensor(offsets[:-1]).cumsum(dim=0)  # 返回维度dim中输入元素的累计和

    return label_list.to(device), text_list.to(device), offsets.to(device)


# 数据加载器
dataloader = DataLoader(train_iter,
                        batch_size=8,
                        shuffle=False,
                        collate_fn=collate_batch)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

3.定义网络模型（简单全连接）

from torch import nn

#这个模型的输入是词汇表大小，嵌入维度，分类种类
# 模型的输出是一个大小为 num_class 的张量，表示每个类别的分数。
class TextClassificationModel(nn.Module):

    def __init__(self, vocab_size, embed_dim, num_class):
        super(TextClassificationModel, self).__init__()

        self.embedding = nn.EmbeddingBag(vocab_size,  # 词典大小
                                         embed_dim,  # 嵌入的维度
                                         sparse=False)  #

        self.fc = nn.Linear(embed_dim, num_class)
        self.init_weights()

    def init_weights(self):
        initrange = 0.5
        self.embedding.weight.data.uniform_(-initrange, initrange)
        self.fc.weight.data.uniform_(-initrange, initrange)
        self.fc.bias.data.zero_()

    def forward(self, text, offsets):
        embedded = self.embedding(text, offsets)
        return self.fc(embedded)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

4.定义参数以及进行训练

num_class  = len(set([label for (label, text) in train_iter]))#利用集合去重的方式等到不同类别的数目
print(num_class)
vocab_size = len(vocab)#计算词汇表大小
em_size     = 64 #指定嵌入维度
model      = TextClassificationModel(vocab_size, em_size, num_class).to(device)

import time
import torch.optim as optim

def train(dataloader):
    model.train()  # 切换为训练模式
    total_acc, train_loss, total_count = 0, 0, 0
    log_interval = 100
    start_time = time.time()

    for idx, (label, text, offsets) in enumerate(dataloader):

        predicted_label = model(text, offsets)

        optimizer.zero_grad()  # grad属性归零
        loss = criterion(predicted_label, label)  # 计算网络输出和真实值之间的差距，label为真实值
        loss.backward()  # 反向传播
        optimizer.step()  # 每一步自动更新

        # 记录acc与loss
        total_acc += (predicted_label.argmax(1) == label).sum().item()
        train_loss += loss.item()
        total_count += label.size(0)

        if idx % log_interval == 0 and idx > 0:
            elapsed = time.time() - start_time
            print('| epoch {:1d} | {:4d}/{:4d} batches '
                  '| train_acc {:4.3f} train_loss {:4.5f}'.format(epoch, idx, len(dataloader),
                                                                  total_acc / total_count, train_loss / total_count))
            total_acc, train_loss, total_count = 0, 0, 0
            start_time = time.time()


def evaluate(dataloader):
    model.eval()  # 切换为测试模式
    total_acc, train_loss, total_count = 0, 0, 0

    with torch.no_grad():
        for idx, (label, text, offsets) in enumerate(dataloader):
            predicted_label = model(text, offsets)

            loss = criterion(predicted_label, label)  # 计算loss值
            # 记录测试数据
            total_acc += (predicted_label.argmax(1) == label).sum().item()
            train_loss += loss.item()
            total_count += label.size(0)

    return total_acc / total_count, train_loss / total_count


from torch.utils.data.dataset import random_split
from torchtext.data.functional import to_map_style_dataset

# 超参数
EPOCHS = 100  # epoch
LR = 3  # 学习率
BATCH_SIZE = 512  # batch size for training

criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=LR)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, 1.0, gamma=0.1)
total_accu = None

train_iter, test_iter = AG_NEWS()  # 加载数据
train_dataset = to_map_style_dataset(train_iter)
test_dataset = to_map_style_dataset(test_iter)
num_train = int(len(train_dataset) * 0.95)

split_train_, split_valid_ = random_split(train_dataset,
                                          [num_train, len(train_dataset) - num_train])

train_dataloader = DataLoader(split_train_, batch_size=BATCH_SIZE,
                              shuffle=True, collate_fn=collate_batch)
valid_dataloader = DataLoader(split_valid_, batch_size=BATCH_SIZE,
                              shuffle=True, collate_fn=collate_batch)
test_dataloader = DataLoader(test_dataset, batch_size=BATCH_SIZE,
                             shuffle=True, collate_fn=collate_batch)

for epoch in range(1, EPOCHS + 1):
    epoch_start_time = time.time()
    train(train_dataloader)
    val_acc, val_loss = evaluate(valid_dataloader)

    if total_accu is not None and total_accu > val_acc:
        scheduler.step()
    else:
        total_accu = val_acc
    print('-' * 69)
    print('| epoch {:1d} | time: {:4.2f}s | '
          'valid_acc {:4.3f} valid_loss {:4.3f}'.format(epoch,
                                                        time.time() - epoch_start_time,
                                                        val_acc, val_loss))

    print('-' * 69)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99

输出结果

| epoch 1 |  100/ 223 batches | train_acc 0.449 train_loss 0.00251
| epoch 1 |  200/ 223 batches | train_acc 0.646 train_loss 0.00195
---------------------------------------------------------------------
| epoch 1 | time: 9.22s | valid_acc 0.727 valid_loss 0.002
---------------------------------------------------------------------
| epoch 2 |  100/ 223 batches | train_acc 0.767 train_loss 0.00136
| epoch 2 |  200/ 223 batches | train_acc 0.815 train_loss 0.00111
---------------------------------------------------------------------
| epoch 2 | time: 9.06s | valid_acc 0.821 valid_loss 0.001
---------------------------------------------------------------------
| epoch 3 |  100/ 223 batches | train_acc 0.842 train_loss 0.00094
| epoch 3 |  200/ 223 batches | train_acc 0.855 train_loss 0.00086
---------------------------------------------------------------------
| epoch 3 | time: 9.06s | valid_acc 0.854 valid_loss 0.001
---------------------------------------------------------------------
| epoch 4 |  100/ 223 batches | train_acc 0.867 train_loss 0.00080
| epoch 4 |  200/ 223 batches | train_acc 0.876 train_loss 0.00075
---------------------------------------------------------------------
| epoch 4 | time: 9.29s | valid_acc 0.868 valid_loss 0.001

**......**

| epoch 98 |  100/ 223 batches | train_acc 0.913 train_loss 0.00052
| epoch 98 |  200/ 223 batches | train_acc 0.912 train_loss 0.00052
---------------------------------------------------------------------
| epoch 98 | time: 9.41s | valid_acc 0.896 valid_loss 0.001
---------------------------------------------------------------------
| epoch 99 |  100/ 223 batches | train_acc 0.914 train_loss 0.00052
| epoch 99 |  200/ 223 batches | train_acc 0.913 train_loss 0.00053
---------------------------------------------------------------------
| epoch 99 | time: 9.74s | valid_acc 0.896 valid_loss 0.001
---------------------------------------------------------------------
| epoch 100 |  100/ 223 batches | train_acc 0.913 train_loss 0.00052
| epoch 100 |  200/ 223 batches | train_acc 0.912 train_loss 0.00053
---------------------------------------------------------------------
| epoch 100 | time: 9.76s | valid_acc 0.896 valid_loss 0.001
---------------------------------------------------------------------

Process finished with exit code 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

可见经过简单的训练可以达到约90%准确率，后续可以继续调参或者调整模型网络进行增进。