五屏网站建设如何,赣州市城乡建设局网站,如何不备案做购物网站,有合作社做网站得不文本分类类似于图片分类#xff0c;也是很常见的一种分类任务#xff0c;将一段不定长的文本序列变换为文本的类别。这节主要就是关注文本的情感分析(sentiment analysis)#xff0c;对电影的评论进行一个正面情绪与负面情绪的分类。整理数据集第一步都是将数据集整理好也是很常见的一种分类任务将一段不定长的文本序列变换为文本的类别。这节主要就是关注文本的情感分析(sentiment analysis)对电影的评论进行一个正面情绪与负面情绪的分类。整理数据集第一步都是将数据集整理好这里我们使用大型电影评论数据集LMDB(Large Movie Review Dataset v1.0)该数据集包含电影评论及其相关二进制情感标签。标签的整体分布是平衡的一半的正类标签和一半的负类标签另外有一些未贴标签的用于无监督学习。电影评分满分是10分将评分7分的判定为正面评论评论得分 4分则为负面评论。下载数据集可以使用自带的函数import d2lzh as d2l
d2l.download_imdb(data_dirdata)或者手动下载http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz自动下载虽然只有80M的大小但是下载特别慢。这里依然推荐迅雷下载下载下来之后就手动解压(自动下载的函数包括自动解压)我们先来看下这个数据集里面有一些什么内容本人地址截图如下可以看到有train和test两个数据集里面都有neg和pos的评论分别表示负面和正面的评论每个文本是一条影评文本名称构造id_评分比如上面图中的200_8.txt表示id为200的这条影评的评分是8分。还有一种feat文件如下图这种.feat文件的格式为LIBSVM是一种用于标记的ascii稀疏向量格式数据比如图片中红色划线处的第200条评论8后面的数字表示什么意思呢8 0:5 1:2 3:1 4:2 6:4 7:7 8:4 9:2 10:2 11:3 16:1 17:3 ... ...这里的0:5表示第一个单词出现了5次1:2就是第二个单词出现了2次后面依次类推。接下来使用自带的read_imdb函数来读取训练集和测试集当然这里使用自带的函数需要注意目录的位置将aclImdb整个目录剪切到上级目录data里面比如本人电脑上的地址D:\data\aclImdbtrain_data, test_data d2l.read_imdb(train), d2l.read_imdb(test)
print(train_data[1])(pygpu) D:\DOG-BREEDpython test.py
[i went to this movie expecting an artsy scary film. what i got was scare after scare. its a horror film at its core. its not dull like other horror films where a haunted house just has ghosts and gore. this film doesnt even show you the majority of the deaths it shows the fear of the characters. i think one of the best things about the concept where its not just the house thats haunted its whoever goes into the house. they become haunted no matter where they are. office buildings, police stations, hotel rooms... etc. after reading some of the external reviews i am really surprised that critics didnt like this film. i am going to see it again this week and am excited about it.br /br /i gave this film 10 stars because it did what a horror film should. it scared the s**t out of me., 1]
返回的结果是列表里面元素是评论加一个正负类标签。这里是赞叹这部恐怖片拍的很不错后面的1表示正类评价。上面两个函数的源码附上[../envs/pygpu/Lib/site-packages/d2lzh/utils.py]def download_imdb(data_dir../data):Download the IMDB data set for sentiment analysis.url (http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz)sha1 01ada507287d82875905620988597833ad4e0903fname gutils.download(url, data_dir, sha1_hashsha1)with tarfile.open(fname, r) as f:f.extractall(data_dir)def read_imdb(foldertrain):Read the IMDB data set for sentiment analysis.data []for label in [pos, neg]:folder_name os.path.join(../data/aclImdb/, folder, label)for file in os.listdir(folder_name):with open(os.path.join(folder_name, file), rb) as f:review f.read().decode(utf-8).replace(\n, ).lower()data.append([review, 1 if label pos else 0])random.shuffle(data)return data预处理数据集数据集和测试集读取没有问题之后我们对评论进行分词这里基于空格分词也是自带的函数get_tokenized_imdb进行分词并做了小写处理。def get_tokenized_imdb(data):Get the tokenized IMDB data set for sentiment analysis.def tokenizer(text):return [tok.lower() for tok in text.split( )]return [tokenizer(review) for review, _ in data]然后将分好词的训练数据集创建Vocabulary词典我们这里过滤掉出现次数少于5的词min_freq5。def get_vocab_imdb(data):Get the vocab for the IMDB data set for sentiment analysis.tokenized_data get_tokenized_imdb(data)counter collections.Counter([tk for st in tokenized_data for tk in st])return text.vocab.Vocabulary(counter, min_freq5)tokenized_data d2l.get_tokenized_imdb(train_data)
vocabd2l.get_vocab_imdb(train_data)
print(len(vocab))#46151可以看到过滤掉次数少的之后词汇量从25000降低到了46151这里返回的变量vocab是mxnet.contrib.text.vocab.Vocabulary类型我们可以查看它里面有哪些属性与方法dir(mxnet.contrib.text.vocab.Vocabulary)[__class__, __delattr__, __dict__, __dir__, __doc__, __eq__, __format__, __ge__, __getattribute__, __gt__, __hash__, __init__, __init_subclass__, __le__, __len__, __lt__, __module__, __ne__, __new__, __reduce__, __reduce_ex__, __repr__, __setattr__, __sizeof__, __str__, __subclasshook__, __weakref__, _index_counter_keys, _index_unknown_and_reserved_tokens, idx_to_token, reserved_tokens, to_indices, to_tokens, token_to_idx, unknown_token]print(vocab.idx_to_token[1])#the由于每条评论的字数或说长度不一样所以不能直接组合成小批量我们通过一个辅助函数让它的长度固定在500超出的进行截断不足的进行pad补足。这个函数preprocess_imdb在d2lzh包中也自带有features, labels d2l.preprocess_imdb(train_data, vocab)
print(features.shape, labels.shape)#(25000, 500) (25000,)从形状可以看到每条评论都固定到了长度为500print(features)[[5.0000e00 5.3200e02 0.0000e00 ... 0.0000e00 0.0000e00 0.0000e00][2.0100e02 5.4810e03 4.2891e04 ... 1.6000e01 2.9200e02 1.1000e01][0.0000e00 0.0000e00 3.6000e01 ... 0.0000e00 0.0000e00 0.0000e00]...[9.0000e00 2.2600e02 3.0000e00 ... 0.0000e00 0.0000e00 0.0000e00][2.8690e03 1.2220e03 1.4000e01 ... 1.1538e04 5.2700e02 2.9000e01][9.0000e00 1.9900e02 1.2108e04 ... 0.0000e00 0.0000e00 0.0000e00]]
NDArray 25000x500 cpu(0)
附上源码def preprocess_imdb(data, vocab):Preprocess the IMDB data set for sentiment analysis.max_l 500def pad(x):return x[:max_l] if len(x) max_l else x [0] * (max_l - len(x))tokenized_data get_tokenized_imdb(data)features nd.array([pad(vocab.to_indices(x)) for x in tokenized_data])labels nd.array([score for _, score in data])return features, labels当然如果想要查看pad对应的值print(vocab.token_to_idx[pad])会报错Traceback (most recent call last): File test.py, line 19, in module print(vocab.token_to_idx[pad])KeyError: pad所以在创建词典Vocabulary的时候需指定参数reserved_tokens[pad]保留这个词def get_vocab_imdb(data):Get the vocab for the IMDB data set for sentiment analysis.tokenized_data d2l.get_tokenized_imdb(data)counter collections.Counter([tk for st in tokenized_data for tk in st])return text.vocab.Vocabulary(counter, min_freq5,reserved_tokens[pad])创建数据迭代器数据集都整理好了之后就开始做数据迭代器每次迭代将返回一个小批量的数据batch_size 64
#train_set gdata.ArrayDataset(*d2l.preprocess_imdb(train_data, vocab))
train_setgdata.ArrayDataset(*[features,labels])
test_set gdata.ArrayDataset(*d2l.preprocess_imdb(test_data, vocab))
train_iter gdata.DataLoader(train_set, batch_size, shuffleTrue)
test_ieter gdata.DataLoader(test_set, batch_size)print(len(train_iter))
for X,y in train_iter:print(X.shape,y.shape)break391
(64, 500) (64,)
创建RNN模型数据迭代器测试没有问题之后接下来就是选择循环神经网络模型来试下效果怎么样了。首先就是将每个词做嵌入也就是通过嵌入层得到特征向量然后我们使用双向循环神经网络对特征序列进一步编码得到序列信息最后将编码的序列信息通过全连接层变换成输出。具体来说我们可以将双向长短期记忆在最初时间步和最终时间步的隐藏状态连结作为特征序列的表征传递给输出层分类。在下面实现BiRNN类中Embedding实例就是嵌入层LSTM实例即为序列编码的隐藏层Dense实例即生成分类结果的输出层。class BiRNN(nn.Block):def __init__(self, vocab, embed_size, num_hiddens, num_layers, **kwargs):super(BiRNN, self).__init__(**kwargs)# 词嵌入层self.embedding nn.Embedding(input_dimlen(vocab), output_dimembed_size)# bidirectional设为True就是双向循环神经网络self.encoder rnn.LSTM(hidden_sizenum_hiddens,num_layersnum_layers,bidirectionalTrue,input_sizeembed_size,)self.decoder nn.Dense(2)def forward(self, inputs):# LSTM需要序列长度(词数)作为第一维所以inputs[形状为(批量大小,词数)]需做转置embeddings self.embedding(inputs.T)print(embeddings.shape)outputs self.encoder(embeddings)print(outputs.shape)# 将初始时间步和最终时间步的隐藏状态作为全连接层输入encoding nd.concat(outputs[0], outputs[-1])print(encoding.shape)outs self.decoder(encoding)return outs# 创建一个含2个隐藏层的双向循环神经网络
embed_size, num_hiddens, num_layers, ctx 100, 100, 2, d2l.try_all_gpus()
net BiRNN(vocabvocab, embed_sizeembed_size, num_hiddensnum_hiddens, num_layersnum_layers
)
net.initialize(init.Xavier(), ctxctx)
#print(net)BiRNN((embedding): Embedding(46152 - 100, float32)(encoder): LSTM(100 - 100, TNC, num_layers2, bidirectional)(decoder): Dense(None - 2, linear)
)
其中LSTM长短期记忆的公式如下(来自源码)训练模型由于情感分类的训练数据集并不大容易过拟合所以这里将使用glove.6B.100d.txt的语料库将这个预训练的词向量作为每个词的特征向量。需要注意的是这里选择的预训练词向量维度是100需要跟创建的模型中的嵌入层输出层大小embed_size一致以及在训练中就不再需要更新这些词向量。glove_embedding text.embedding.create(glove, pretrained_file_nameglove.6B.100d.txt, vocabularyvocab
)
net.embedding.weight.set_data(glove_embedding.idx_to_vec)
net.embedding.collect_params().setattr(grad_req,null)lr,num_epochs0.01,5
trainergluon.Trainer(net.collect_params(),adam,{learning_rate:lr})
lossgloss.SoftmaxCrossEntropyLoss()
d2l.train(train_iter,test_ieter,net,loss,trainer,ctx,num_epochs)print(d2l.predict_sentiment(net,vocab,[this,movie,is,so,good]))
print(d2l.predict_sentiment(net,vocab,[this,movie,is,so,bad]))training on [gpu(0)]
epoch 1, loss 0.6553, train acc 0.605, test acc 0.738, time 65.4 sec
epoch 2, loss 0.4273, train acc 0.807, test acc 0.809, time 65.4 sec
epoch 3, loss 0.3514, train acc 0.851, test acc 0.849, time 65.5 sec
epoch 4, loss 0.3054, train acc 0.874, test acc 0.859, time 65.6 sec
epoch 5, loss 0.2765, train acc 0.887, test acc 0.843, time 65.6 sec
positive
negative
其中预测函数的源码如下def predict_sentiment(net, vocab, sentence):Predict the sentiment of a given sentence.sentence nd.array(vocab.to_indices(sentence), ctxtry_gpu())label nd.argmax(net(sentence.reshape((1, -1))), axis1)return positive if label.asscalar() 1 else negative
