[人工智能]文本分类之textCNN

近期利用零散的时间，系统的学习下 nlp经典的模型，主要是看了韩国一个小哥哥的github开源代码，写的很好，大家有时间 可以去看看，美中不足的是没有注释和说明，这里我主要是参考他的代码，自己学习学习，并顺便把nlp的零散的基础知识补充一下

基础知识回顾：

一. CNN网络

二. textCNN网络

三.代码展示:

import torch
import torch.nn as nn
import torch.optim as optim
import pdb
import torch.nn.functional as F
import numpy as np

### textCNN 文本分类
num_class = 2 ## 二分类
embedding_size = 3

class TextCNN(nn.Module):
    def __init__(self,num_channels = 1) -> None:  ## input  [6,1,3,3]
        super(TextCNN,self).__init__()
        
        self.embed = nn.Embedding(voc_size, embedding_size)
        self.conv1 = nn.Conv2d(num_channels,5,1,1) # num_channel = 1 表示通道数, 5 表示 filter kernel 的数量，滤波器的数量
        ###  1 -> filter的大小 1*1 的大小, 1 -> stride 步子的大小
        self.conv2 = nn.Conv2d(5,5,1,1)  ## input = [6,5,3,3] batch = 6, channel_num = 5,filter_size =1, stride = 1
        self.fc = nn.Linear(5*3*3,num_class)

    def forward(self,X):
        
        embedded = self.embed(X)  ##  X [6,3],embedded = [6,3,3]

        X = embedded.unsqueeze(1)   ## X = [6,1,3,3]

        out1 = self.conv1(X) ## X -> [6,1,3,3]   out1 -> [6,5,3,3]   3 = (3-1)/stride + 1 
        y1 = F.relu(out1)  ##  y -> [6,5,3,3]
        y1 = F.max_pool2d(y1,1,1)

        out2 = self.conv2(y1) ## out2 = [6,5,3,3]
        y2 = F.relu(out2)  ##  y -> [6,5,3,3]

        y2 = F.max_pool2d(y2,1,1)

        y2 = y2.view(-1,5*3*3)
        out = self.fc(y2)

        return out        

if __name__ == '__main__':

    sentences = ["i love you", "he loves me", "she likes baseball", "i hate you", "sorry for that", "this is awful"]
    labels = [1, 1, 1, 0, 0, 0]

    word_list = ' '.join(sentences).split()
    word_list = list(set(word_list))
    word_dict = {w:i for i,w in enumerate(word_list)}
    voc_size = len(word_list)


    model = TextCNN()

    criterion = nn.CrossEntropyLoss()
    optimizer = optim.Adam(model.parameters(), lr=0.001)

    inputs = torch.LongTensor([np.asarray([word_dict[n] for n in sen.split()]) for sen in sentences])
    targets = torch.LongTensor([out for out in labels]) # To using Torch Softmax Loss function

    # pdb.set_trace()
    # Training
    for epoch in range(5000):
        optimizer.zero_grad()
        output = model(inputs)

        # output : [batch_size, num_classes], target_batch : [batch_size] (LongTensor, not one-hot)
        loss = criterion(output, targets)
        if (epoch + 1) % 200 == 0:
            print('Epoch:', '%04d' % (epoch + 1), 'cost =', '{:.6f}'.format(loss))

        loss.backward()
        optimizer.step()

    # Test
    test_text = 'sorry hate you'
    tests = [np.asarray([word_dict[n] for n in test_text.split()])]
    test_batch = torch.LongTensor(tests)

    # Predict
    predict = model(test_batch).data.max(1, keepdim=True)[1]
    if predict[0][0] == 0:
        print(test_text,"is Bad Mean...")
    else:
        print(test_text,"is Good Mean!!")