PyTorch LSTM的一个简单例子:实现单词词性判断

      本文将使用LSTM来判别一句话中每一个单词的词性。在一句话中,如果我们孤立地看某一个单词,比如单词book,而不看book前面的单词,就不能准确的判断book在这句话中是动词还是名词,但如果我们能记住book前面出现的单词,那么就能很有把握地判断book的词性。LSTM神经网络就能记住前面的单词。关于LSTM的详细介绍,大家可参考文末的参考资料[1][2]。

      下面的代码主要来自文末的参考资料[3],本文对原代码做了修改并增加了注释,使其变得更简单易懂。要理解下面的程序,理解torch.nn.Embedding是关键之一,这篇博客将提供帮助。

'''
本程序实现了对单词词性的判断,输入一句话,输出该句话中每个单词的词性。
'''

import torch
import torch.nn.functional as F
from torch import nn, optim

training_data = [("The dog ate the apple".split(), ["DET", "NN", "V", "DET", "NN"]),
                 ("Everybody read that book".split(), ["NN", "V", "DET", "NN"])]

word_to_idx = {}
tag_to_idx = {}
for context, tag in training_data:
    for word in context:
        if word not in word_to_idx:
            word_to_idx[word] = len(word_to_idx)
    for label in tag:
        if label not in tag_to_idx:
            tag_to_idx[label] = len(tag_to_idx)
idx_to_tag = {tag_to_idx[tag]: tag for tag in tag_to_idx}


class LSTMTagger(nn.Module):
    def __init__(self, n_word, n_dim, n_hidden, n_tag):
        super(LSTMTagger, self).__init__()
        self.word_embedding = nn.Embedding(n_word, n_dim)
        self.lstm = nn.LSTM(n_dim, n_hidden, batch_first=True)  # nn.lstm()接受的数据输入是(序列长度,batch,输入维数),
                                                                # 这和我们cnn输入的方式不太一致,所以使用batch_first=True,把输入变成(batch,序列长度,输入维度),本程序的序列长度指的是一句话的单词数目
                                                                # 同时,batch_first=True会改变输出的维度顺序。
self.linear1 = nn.Linear(n_hidden, n_tag) def forward(self, x): # x是word_list,即单词的索引列表,size为len(x) x = self.word_embedding(x) # embedding之后,x的size为(len(x),n_dim) x = x.unsqueeze(0) # unsqueeze之后,x的size为(1,len(x),n_dim),1在下一行程序的lstm中被当做是batchsize,len(x)被当做序列长度 x, _ = self.lstm(x) # lstm的隐藏层输出,x的size为(1,len(x),n_hidden),因为定义lstm网络时用了batch_first=True,所以1在第一维,如果batch_first=False,则len(x)会在第一维 x = x.squeeze(0) # squeeze之后,x的size为(len(x),n_hidden),在下一行的linear层中,len(x)被当做是batchsize x = self.linear1(x) # linear层之后,x的size为(len(x),n_tag) y = F.log_softmax(x, dim=1) # 对第1维先进行softmax计算,然后log一下。y的size为(len(x),n_tag)。 return y model = LSTMTagger(len(word_to_idx), 100, 128, len(tag_to_idx)) if torch.cuda.is_available(): model = model.cuda() criterion = nn.NLLLoss() optimizer = optim.SGD(model.parameters(), lr=1e-2) for epoch in range(200): running_loss = 0 for data in training_data: sentence, tags = data word_list = [word_to_idx[word] for word in sentence] # word_list是word索引列表 word_list = torch.LongTensor(word_list) tag_list = [tag_to_idx[tag] for tag in tags] # tag_list是tag索引列表 tag_list = torch.LongTensor(tag_list) if torch.cuda.is_available(): word_list = word_list.cuda() tag_list = tag_list.cuda() # forward out = model(word_list) loss = criterion(out, tag_list) running_loss += loss.data.numpy() # backward optimizer.zero_grad() loss.backward() optimizer.step() print('Epoch: {:<3d} | Loss: {:6.4f}'.format(epoch, running_loss / len(data))) # 模型测试 test_sentence = "Everybody ate the apple" print(' The test sentence is: ', test_sentence) test_sentence = test_sentence.split() test_list = [word_to_idx[word] for word in test_sentence] test_list = torch.LongTensor(test_list) if torch.cuda.is_available(): test_list = test_list.cuda() out = model(test_list) _, predict_idx = torch.max(out, 1) # 1表示找行的最大值。 predict_idx是词性索引,是一个size为([len(test_sentence)]的张量 predict_tag = [idx_to_tag[idx] for idx in list(predict_idx.numpy())] print('The predict tags are:', predict_tag)

参考资料:

[1] 零基础入门深度学习(6) - 长短时记忆网络(LSTM)

[2] 10分钟快速入门PyTorch (5)

[3] 10分钟快速入门PyTorch (9)

原文地址:https://www.cnblogs.com/picassooo/p/12544439.html