这个系列一共有8个部分。主要参考了github上的几个代码。
使用工具有torchtext，pytorch。
数据集主要是烂番茄电影评论数据集https://www.kaggle.com/c/sentiment-analysis-on-movie-reviews。
烂番茄电影评论数据集是包含了训练集和测试集，原始的句子已经通过斯坦福大学的解析器解析成为了很多短句，是数据集中的phrase ，另外提供了sentenceid可以寻找到原始的句子，情绪标签分为五个
0 - 消极
1 - 有些消极
2 - 中立
3 - 有点积极
4 - 积极
首先先观察数据
1 import pandas as pd
2
3 files1 = pd.read_csv("datasets/train.tsv",sep="\t")
4 files2 = pd.read_csv("datasets/test.tsv",sep="\t")
5 files1.head(5)
6 files2.head(5) train_data
test_data
利用torchtext构建数据集
import torchtext
from torchtext import data,datasets
train_data = data.TabularDataset(
path='./datasets/train.tsv',format='tsv',
skip_header=True,
fields = [('PhraseId', None),('SentenceId', None),('Phrase', TEXT),('Sentiment',LABEL)])
test_data = data.TabularDataset(
path='./datasets/test.tsv',format='tsv',
skip_header=True,
fields = [('PhraseId', None),('SentenceId', None),('Phrase', TEXT)])import random
train_data, valid_data = train_data.split(random_state = random.seed(10))
构造词典和batch
from torchtext.vocab import GloVe
from torchtext import data
TEXT.build_vocab(train_data, vectors=GloVe(name='6B', dim=300))
LABEL.build_vocab()
BATCH_SIZE = 64
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') train_iterator, valid_iterator, test_iterator = data.BucketIterator.splits( (train_data, valid_data, test_data), batch_size = BATCH_SIZE, device = device)
使用最简单的RNN，如下
import torch.nn as nn
class RNN(nn.Module):
def __init__(self, vocab_size, embedding_dim, hidden_dim, output_dim):
super().__init__()
self.embedding = nn.Embedding(vocab_size, embedding_dim)
self.rnn = nn.RNN(embedding_dim, hidden_dim,frist_batch=True)
self.fc = nn.Linear(hidden_dim, output_dim)
def forward(self, text):
embedded = self.embedding(text).permute(1,0,2)
output, hidden = self.rnn(embedded)
return self.fc(hidden)
import torch.optim as optim
import torch.nn.functional as F
model = RNN()
model.train()
optimizer = optim.Adam(model.parameters(), lr=0.01)
loss_funtion = F.cross_entropy
for epoch in range(2):
for batch in train_iterator:
optimizer.zero_grad()
predicted = model(batch.Phrase)
loss = loss_funtion(predicted, batch.Sentiment)
loss.backward()
optimizer.step()
不怎么会写，后面补充完。。。为啥别人的博客写的那么好！！！
转载于:https://www.cnblogs.com/laohejb/p/11131601.html