深度学习笔记40_中文文本分类-Pytorch实现

L10732034822025-05-03 21:03

一、我的环境

1.语言环境：Python 3.8

2.编译器：Pycharm

3.深度学习环境：

  • torch==1.12.1+cu113
  • torchvision==0.13.1+cu113

二**、导入数据**

复制代码 
import torch
import torch.nn as nn
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warnings
 
warnings.filterwarnings("ignore")             #忽略警告信息
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

import pandas as pd
 
# 加载自定义中文数据
train_data = pd.read_csv('./data/train.csv', sep='\t', header=None)
print(train_data.head())

结果：

复制代码 
                       0              1
0      还有双鸭山到淮阴的汽车票吗13号的   Travel-Query
1                从这里怎么回家   Travel-Query
2       随便播放一首专辑阁楼里的佛里的歌     Music-Play
3              给看一下墓王之王嘛  FilmTele-Play
4  我想看挑战两把s686打突变团竞的游戏视频     Video-Play

三**、构建词典**

复制代码 
from torchtext.data.utils import get_tokenizer
from torchtext.vocab import build_vocab_from_iterator
import jieba
 
# 中文分词方法
tokenizer = jieba.lcut
 
def yield_tokens(data_iter):
    for text,_ in data_iter:
        yield tokenizer(text)
 
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(['我','想','看','和平','精英','上','战神','必备','技巧','的','游戏','视频']))

结果：[2, 10, 13, 973, 1079, 146, 7724, 7574, 7793, 1, 186, 28]

复制代码 
text_pipeline  = lambda x: vocab(tokenizer(x))
label_pipeline = lambda x: label_name.index(x)
 
print(text_pipeline('我想看和平精英上战神必备技巧的游戏视频'))
print(label_pipeline('Video-Play'))

结果：[2, 10, 13, 973, 1079, 146, 7724, 7574, 7793, 1, 186, 28]
4

四**、**生成数据批次和迭代器

复制代码 
from torch.utils.data import DataLoader
 
 
def collate_batch(batch):
    label_list, text_list, offsets = [], [], [0]
 
    for (_text, _label) 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 text_list.to(device), label_list.to(device), offsets.to(device)
 
 
# 数据加载器，调用示例
dataloader = DataLoader(train_iter,
                        batch_size=8,
                        shuffle=False,
                        collate_fn=collate_batch)

五**、**定义模型

复制代码 
from torch import nn
 
 
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)  # 初始化权重
        self.fc.weight.data.uniform_(-initrange, initrange)
        self.fc.bias.data.zero_()  # 偏置值归零
 
    def forward(self, text, offsets):
        embedded = self.embedding(text, offsets)
        return self.fc(embedded)

六**、**定义实例

复制代码 
num_class  = len(label_name)
vocab_size = len(vocab)
em_size    = 64
model      = TextClassificationModel(vocab_size, em_size, num_class).to(device)

七**、**定义训练函数与评估函数

复制代码 
import time
 
def train(dataloader):
    model.train()  # 切换为训练模式
    total_acc, train_loss, total_count = 0, 0, 0
    log_interval = 50
    start_time = time.time()
 
    for idx, (text, label, 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
        torch.nn.utils.clip_grad_norm_(model.parameters(), 0.1)  # 梯度裁剪
        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, (text, label, 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

八**、**训练模型

复制代码 
from torch.utils.data.dataset import random_split
from torchtext.data.functional import to_map_style_dataset
 
# 超参数
EPOCHS = 10  # epoch
LR = 5  # 学习率
BATCH_SIZE = 64  # 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
 
# 构建数据集
train_iter = coustom_data_iter(train_data[0].values[:], train_data[1].values[:])
train_dataset = to_map_style_dataset(train_iter)
 
split_train_, split_valid_ = random_split(train_dataset,
                                          [int(len(train_dataset) * 0.8), int(len(train_dataset) * 0.2)])
 
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)
 
for epoch in range(1, EPOCHS + 1):
    epoch_start_time = time.time()
    train(train_dataloader)
    val_acc, val_loss = evaluate(valid_dataloader)
 
    # 获取当前的学习率
    lr = optimizer.state_dict()['param_groups'][0]['lr']
 
    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} | lr {:4.6f}'.format(epoch,
                                                                     time.time() - epoch_start_time,
                                                                     val_acc, val_loss, lr))
 
    print('-' * 69)

结果：

复制代码 
Batch [50/152], Loss: 0.0340, Accuracy: 0.4203
Batch [100/152], Loss: 0.0235, Accuracy: 0.5851
Batch [150/152], Loss: 0.0309, Accuracy: 0.6572
---------------------------------------------------------------------
| epoch 1 | time: 0.55s | valid_acc 0.814 valid_loss 0.012 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0104, Accuracy: 0.8165
Batch [100/152], Loss: 0.0099, Accuracy: 0.8215
Batch [150/152], Loss: 0.0092, Accuracy: 0.8329
---------------------------------------------------------------------
| epoch 2 | time: 0.44s | valid_acc 0.855 valid_loss 0.008 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0068, Accuracy: 0.8790
Batch [100/152], Loss: 0.0065, Accuracy: 0.8778
Batch [150/152], Loss: 0.0064, Accuracy: 0.8809
---------------------------------------------------------------------
| epoch 3 | time: 0.44s | valid_acc 0.874 valid_loss 0.007 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0050, Accuracy: 0.9105
Batch [100/152], Loss: 0.0051, Accuracy: 0.9101
Batch [150/152], Loss: 0.0048, Accuracy: 0.9130
---------------------------------------------------------------------
| epoch 4 | time: 0.44s | valid_acc 0.882 valid_loss 0.006 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0039, Accuracy: 0.9366
Batch [100/152], Loss: 0.0039, Accuracy: 0.9339
Batch [150/152], Loss: 0.0038, Accuracy: 0.9350
---------------------------------------------------------------------
| epoch 5 | time: 0.44s | valid_acc 0.896 valid_loss 0.006 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0028, Accuracy: 0.9519
Batch [100/152], Loss: 0.0030, Accuracy: 0.9517
Batch [150/152], Loss: 0.0030, Accuracy: 0.9494
---------------------------------------------------------------------
| epoch 6 | time: 0.44s | valid_acc 0.898 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0025, Accuracy: 0.9580
Batch [100/152], Loss: 0.0024, Accuracy: 0.9616
Batch [150/152], Loss: 0.0024, Accuracy: 0.9609
---------------------------------------------------------------------
| epoch 7 | time: 0.44s | valid_acc 0.902 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0018, Accuracy: 0.9764
Batch [100/152], Loss: 0.0019, Accuracy: 0.9739
Batch [150/152], Loss: 0.0019, Accuracy: 0.9724
---------------------------------------------------------------------
| epoch 8 | time: 0.44s | valid_acc 0.900 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0015, Accuracy: 0.9810
Batch [100/152], Loss: 0.0014, Accuracy: 0.9817
Batch [150/152], Loss: 0.0014, Accuracy: 0.9818
---------------------------------------------------------------------
| epoch 9 | time: 0.49s | valid_acc 0.906 valid_loss 0.005 | lr 0.500000
---------------------------------------------------------------------
Batch [50/152], Loss: 0.0013, Accuracy: 0.9831
Batch [100/152], Loss: 0.0013, Accuracy: 0.9831
Batch [150/152], Loss: 0.0014, Accuracy: 0.9825
---------------------------------------------------------------------
| epoch 10 | time: 0.54s | valid_acc 0.906 valid_loss 0.005 | lr 0.500000
---------------------------------------------------------------------

九**、预测**

复制代码 
def predict(text, text_pipeline):
    with torch.no_grad():
        text = torch.tensor(text_pipeline(text))
        output = model(text, torch.tensor([0]))
        return output.argmax(1).item()
 
# ex_text_str = "随便播放一首专辑阁楼里的佛里的歌"
ex_text_str = "还有双鸭山到淮阴的汽车票吗13号的"
 
model = model.to("cpu")
 
print("该文本的类别是：%s" %label_name[predict(ex_text_str, text_pipeline)])

该文本的类别是：Travel-Query

总结：

  1. ‌**语料库（原始文本）**‌：

    来源包括维基百科、网页文本、新闻资讯及内部文本。

  2. ‌文本清洗‌：

    清洗原始文本，包括去除标点符号和特殊字符。该流程主要用于将原始文本数据转化为可用于模型训练的数值化向量，再通过深度学习模型进行文本分类。

    • ‌分词‌：

      使用jieba分词工具对清洗后的文本进行分词处理。

    • ‌建模‌：

      采用不同的模型进行文本建模，包括循环神经网络（RNN）、卷积神经网络（CNN）、门控循环单元（GRU）和长短期记忆网络（LSTM）。

    • ‌文本向量化‌：

      将分词后的文本转换为向量表示，方法包括TF-IDF和Word2vec。

相关推荐
杨浦老苏23 分钟前
Nanote：极简Markdown笔记应用
笔记·docker·markdown·群晖
汇能感知27 分钟前
光谱相机如何提升目标检测与识别精度
经验分享·笔记·科技
踢足球的程序员·1 小时前
GAMES202-高质量实时渲染（Assignment 2）
笔记·学习·图形渲染
大模型铲屎官1 小时前
【深度学习-Day 2】图解线性代数：从标量到张量，理解深度学习的数据表示与运算
人工智能·pytorch·python·深度学习·线性代数·机器学习·llm
蹦蹦跳跳真可爱5891 小时前
Python----卷积神经网络（LeNet-5的手写体识别）
人工智能·python·深度学习·神经网络·cnn
孞㐑￥2 小时前
Linux之基础开发工具(yum,vim,gcc,g++)
linux·c++·经验分享·笔记
Francek Chen2 小时前
【现代深度学习技术】现代循环神经网络04：双向循环神经网络
人工智能·pytorch·rnn·深度学习·神经网络
字节旅行者3 小时前
Matlab自学笔记
开发语言·笔记·matlab
搏博7 小时前
机器学习之五：基于解释的学习
人工智能·深度学习·学习·算法·机器学习
热门推荐
01KGG转MP3工具|非KGM文件|解密音频02西电B测-计算机网络综合实验(含验收问题)03从零安装 LLaMA-Factory 微调 Qwen 大模型成功及所有的坑04【解决】Android Gradle Sync 报错 Could not read workspace metadata05Coze扣子平台完整体验和实践（附国内和国际版对比）06YOLOv8入门 | 重要性能衡量指标、训练结果评价及分析及影响mAP的因素【发论文关注的指标】07AI生成写真，从产品经理的角度看看「妙鸭相机」08DeepSeek各版本说明与优缺点分析09苍穹外卖面试总结10Springboot集成Spring AI和Milvus，验证RAG构建过程