数据集下载地址
包含2000条文本数据，每条文本对应一个情感分类，共有6种情感 love,sadness,fear, surprise, anger, joy。

%pip install nltk

Collecting nltkUsing cached nltk-3.7-py3-none-any.whl (1.5 MB)
Collecting regex>=2021.8.3Using cached regex-2022.7.9-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (763 kB)
Requirement already satisfied: click in /opt/conda/lib/python3.9/site-packages (from nltk) (8.1.2)
Requirement already satisfied: joblib in /opt/conda/lib/python3.9/site-packages (from nltk) (1.1.0)
Requirement already satisfied: tqdm in /opt/conda/lib/python3.9/site-packages (from nltk) (4.63.1)
Installing collected packages: regex, nltk
Successfully installed nltk-3.7 regex-2022.7.9
[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv[0m[33m
[0mNote: you may need to restart the kernel to use updated packages.

import requests
from tqdm import tqdm
import numpy as np
import nltk

样本总数目: 20000

加载数据集

加载数据方式看个人需求，最后构建成句子列表和标签列表即可，且注意句子和标签顺序相呼应

all_sentences, all_label = [], []
for i in tqdm(range(0, NumberOfSamples, 100)):res = requests.get('***************',headers=HEADERS)for data in res.json()["content"]:all_sentences.append(data["data"]["text"])all_label.append(data["annotations"][0]["dataAnnos"]["label"])

100%|██████████| 200/200 [04:46<00:00,  1.43s/it]

构造DataFrame 方便统计

import pandas as pddf = pd.DataFrame(list(zip(all_sentences, all_label)), columns = ['sentence', 'label'])
df.head(10)
sentence label
0 i didnt feel humiliated sadness
1 i can go from feeling so hopeless to so damned... sadness
2 im grabbing a minute to post i feel greedy wrong anger
3 i am ever feeling nostalgic about the fireplac... love
4 i am feeling grouchy anger
5 ive been feeling a little burdened lately wasn... sadness
6 ive been taking or milligrams or times recomme... surprise
7 i feel as confused about life as a teenager or... fear
8 i have been with petronas for years i feel tha... joy
9 i feel romantic too love 
print(df.shape)
print(df.isnull().sum())   # 检查空值

(20000, 2)
sentence    0
label       0
dtype: int64

文本处理, 去除一些无用字符

import re
import string
"""
移除网址，$数字$, 数字， 标点符号
"""
def remove_hyperlinks(text):sentence = re.sub(r'www?:\/\/.*[\r\n]*', '', text)  sentence = re.sub(r'http?:\/\/.*[\r\n]*', '', sentence)return re.sub(r'https?:\/\/.*[\r\n]*', '', sentence)def remove_currencies(text):return re.sub(r'[\$\d+\d+\$]', '', text)def remove_number(text):return re.sub(r'\d+', '', text)def remove_punctuation(text):return ''.join([word for word in text if word not in string.punctuation])df['sentence'] = df['sentence'].apply(lambda x: remove_hyperlinks(x.lower()))
df['sentence'] = df['sentence'].apply(lambda x: remove_currencies(x))
df['sentence'] = df['sentence'].apply(lambda x: remove_number(x))
df.head(10)
sentence label
0 i didnt feel humiliated sadness
1 i can go from feeling so hopeless to so damned... sadness
2 im grabbing a minute to post i feel greedy wrong anger
3 i am ever feeling nostalgic about the fireplac... love
4 i am feeling grouchy anger
5 ive been feeling a little burdened lately wasn... sadness
6 ive been taking or milligrams or times recomme... surprise
7 i feel as confused about life as a teenager or... fear
8 i have been with petronas for years i feel tha... joy
9 i feel romantic too love 

# 根据空格获取单词
def tokenize(text):return re.split(r'\W+', text) # 大写W 除数字字母下划线之外的字符, 在这里主要是空格df['sentence'] = df['sentence'].apply(lambda x: remove_punctuation(x))
df['sentence'] = df['sentence'].apply(lambda x: tokenize(x))
df.head(10)
sentence label
0 [i, didnt, feel, humiliated] sadness
1 [i, can, go, from, feeling, so, hopeless, to, ... sadness
2 [im, grabbing, a, minute, to, post, i, feel, g... anger
3 [i, am, ever, feeling, nostalgic, about, the, ... love
4 [i, am, feeling, grouchy] anger
5 [ive, been, feeling, a, little, burdened, late... sadness
6 [ive, been, taking, or, milligrams, or, times,... surprise
7 [i, feel, as, confused, about, life, as, a, te... fear
8 [i, have, been, with, petronas, for, years, i,... joy
9 [i, feel, romantic, too] love

去除停用词

from nltk.corpus import stopwords
nltk.download('stopwords')stopword = stopwords.words('english')
def remove_stopword(text):return [word for word in text if word not in stopword]df['sentence'] = df['sentence'].apply(lambda x: remove_stopword(x))
df.head(10)

[nltk_data] Downloading package stopwords to /home/jovyan/nltk_data...
[nltk_data]   Package stopwords is already up-to-date!
sentence label
0 [didnt, feel, humiliated] sadness
1 [go, feeling, hopeless, damned, hopeful, aroun... sadness
2 [im, grabbing, minute, post, feel, greedy, wrong] anger
3 [ever, feeling, nostalgic, fireplace, know, st... love
4 [feeling, grouchy] anger
5 [ive, feeling, little, burdened, lately, wasnt... sadness
6 [ive, taking, milligrams, times, recommended, ... surprise
7 [feel, confused, life, teenager, jaded, year, ... fear
8 [petronas, years, feel, petronas, performed, w... joy
9 [feel, romantic] love 
from nltk.stem import WordNetLemmatizer
nltk.download('wordnet')
nltk.download('omw-1.4')
# 把一些名词的词性还原 如 cars-->car
lemmatizer = WordNetLemmatizer()
def lemmatize(text):return ' '.join([lemmatizer.lemmatize(word) for word in text])df['sentence'] = df['sentence'].apply(lambda x: lemmatize(x))
df.head(10)

[nltk_data] Downloading package wordnet to /home/jovyan/nltk_data...
[nltk_data]   Package wordnet is already up-to-date!
[nltk_data] Downloading package omw-1.4 to /home/jovyan/nltk_data...
[nltk_data]   Package omw-1.4 is already up-to-date!
sentence label
0 didnt feel humiliated sadness
1 go feeling hopeless damned hopeful around some... sadness
2 im grabbing minute post feel greedy wrong anger
3 ever feeling nostalgic fireplace know still pr... love
4 feeling grouchy anger
5 ive feeling little burdened lately wasnt sure sadness
6 ive taking milligram time recommended amount i... surprise
7 feel confused life teenager jaded year old man fear
8 petronas year feel petronas performed well mad... joy
9 feel romantic love 
# 获取类别
category = df["label"].unique().tolist()

建立单词字典

sentence = df['sentence'].values
label = df["label"].apply(lambda x: category.index(x)).values

# 获取所有不重复的单词
words = [word.lower() for s in sentence for word in s.split(" ")]
various_words = list(set(words))

int2word = dict(enumerate(various_words))  # 索引到单词
word2int = {w:int(i) for i,w in int2word.items()} # 单词到索引

from collections import Counter
Counter(label)

Counter({0: 5797, 1: 2709, 2: 1641, 3: 719, 4: 2373, 5: 6761})

清理文本过长或过短的句子

sentence_length = [len(s.split()) for s in sentence]   # 每个句子单词的个数
counts = dict(Counter(sentence_length))

import matplotlib.pyplot as plt
plt.figure(figsize=(16,5))
plt.bar(counts.keys(),counts.values())
plt.xlabel("sentence_length")
plt.ylabel("num")
plt.show()

​

min_sen = min(counts.items())
max_sen = max(counts.items())
print("min:{}, max:{}".format(min_sen,max_sen))

min:(1, 8), max:(35, 2)

"""获取最小和最大长度的句子的索引
"""
min_index = [i for i,length in enumerate(sentence_length) if length==min_sen[0]]
max_index = [i for i,length in enumerate(sentence_length) if length==max_sen[0]]

new_text = np.delete(sentence, min_index)
new_text2 = np.delete(new_text, max_index)
print("原始文本数量：",len(new_text))
print("新文本数量: ", len(new_text2))

原始文本数量： 19992
新文本数量:  19990

new_labels = np.delete(label, min_index)
new_labels = np.delete(new_labels, max_index)
print("原始标签数量：",len(new_text))
print("新标签数量: ", len(new_text2))

原始标签数量： 19992
新标签数量:  19990

将句子根据单词字典映射成数字

text2ints = []
for sentence in new_text2:sample = list()for word in sentence.split():int_value = word2int[word]sample.append(int_value)text2ints.append(sample)

text2ints[:5]

[[776, 2470, 276],[2272, 11006, 8438, 4651, 3985, 4269, 8191, 7923, 2237],[10985, 14024, 775, 12813, 2470, 11920, 11414],[1906, 11006, 12544, 9257, 7443, 5937, 11229],[11006, 11453]]

数据的填充或者截断

def reset_text(text, seq_len):dataset = np.zeros((len(text),seq_len))for index,sentence in enumerate(text):if len(sentence) < seq_len:dataset[index, :len(sentence)] = sentence  # 后面填充0else:dataset[index, :] = sentence[:seq_len]  # 截断return datasetdataset = reset_text(text2ints, seq_len=22)

dataset

array([[  776.,  2470.,   276., ...,     0.,     0.,     0.],[ 2272., 11006.,  8438., ...,     0.,     0.,     0.],[10985., 14024.,   775., ...,     0.,     0.,     0.],...,[ 2470.,   782., 14928., ...,     0.,     0.,     0.],[10985., 11006.,  3284., ...,     0.,     0.,     0.],[ 2470., 13911.,  7535., ...,     0.,     0.,     0.]])

import torch
import torch.nn as nn

## 转化为张量
dataset_tensor = torch.from_numpy(dataset)
label_tensor = torch.from_numpy(new_labels)
print(type(dataset_tensor), type(label_tensor))

<class 'torch.Tensor'> <class 'torch.Tensor'>

划分数据集

all_samples = len(dataset_tensor)
train_ratio = 0.8
val_ratio = 0.2

train = dataset_tensor[:int(train_ratio*all_samples)]
train_labels = label_tensor[:int(train_ratio*all_samples)]val = dataset_tensor[int(train_ratio*all_samples):]
val_labels = label_tensor[int(train_ratio*all_samples):]print("训练集:{}----{}".format(train.shape, train_labels.shape))print("验证集:{}----{}".format(val.shape,val_labels.shape))

训练集:torch.Size([15992, 22])----torch.Size([15992])
验证集:torch.Size([3998, 22])----torch.Size([3998])

封装dataloader

from torch.utils.data import TensorDataset, DataLoadertrain_dataset = TensorDataset(train, train_labels)
val_dataset = TensorDataset(val, val_labels)batch_size = 128train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=8, pin_memory=True, drop_last=True)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=True, num_workers=8, pin_memory=True, drop_last=True)
device = "cuda" if torch.cuda.is_available() else "cpu"
print(device)

cpu

定义网络模型

class SentimentNet(nn.Module):def __init__(self, input_size, embedding_dim, hidden_dim, output_size, num_layers, dropout=0.5):super(SentimentNet, self).__init__()self.hidden_dim = hidden_dimself.output_size = output_sizeself.num_layers = num_layersself.embedding= nn.Embedding(input_size, embedding_dim)self.lstm = nn.LSTM(embedding_dim, hidden_dim, num_layers, dropout=dropout,batch_first=True)self.linear = nn.Linear(hidden_dim, 128)self.relu = nn.ReLU()self.linear2 = nn.Linear(128, output_size)self.dropout = nn.Dropout(p=0.4)def forward(self, x, hidden):"""x:  (128,10)"""batch_size = x.size(0)x = x.long() embeds = self.embedding(x) # embeds(128,10,200)#  out(128,22,128)--batch_size,seq_len, hidden_size#  hidden是个元组{h_n,c_n}每个都是(2, 128,128) -- num_layer,batch_size, hidden_sizeout,hidden = self.lstm(embeds, hidden) out = self.linear(out[:, -1, :]) # out(128，128)out = self.dropout(out)out = self.relu(out) out = self.linear2(out) # out(128, 6)return out, hidden # hidden (h_n, c_n)def init_hidden(self, batch_size):weight = next(self.parameters())# h_0, c_0return (weight.new_zeros(self.num_layers, batch_size, self.hidden_dim),weight.new_zeros(self.num_layers, batch_size, self.hidden_dim))
input_size = len(word2int)
output_size = len(category)
embedding_dim = 200
hidden_dim= 128
num_layers= 2

model = SentimentNet(input_size, embedding_dim, hidden_dim, output_size, num_layers)
model.to(device)

SentimentNet((embedding): Embedding(15104, 200)(lstm): LSTM(200, 128, num_layers=2, batch_first=True, dropout=0.5)(linear): Linear(in_features=128, out_features=128, bias=True)(relu): ReLU()(linear2): Linear(in_features=128, out_features=6, bias=True)(dropout): Dropout(p=0.4, inplace=False)
)

criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.RMSprop(model.parameters(), lr=1e-3)
# Decay LR by a factor of 0.1 every 3 epochs
exp_lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)

模型训练

def train(model, data_loader, criterion, optimizer, scheduler, num_epochs):train_process = dict()train_loss_epoch10, val_loss_epoch10= [],[]val_acc_epoch10 = []for epoch in range(num_epochs):hs = model.init_hidden(batch_size)train_loss = []train_correct = 0model.train()for data, target in data_loader:  data = data.to(device)target = target.to(device)output,hs = model(data, hs)preds = torch.argmax(output, dim=1)train_correct += torch.sum(preds==target)hs = tuple([h.data for h in hs])loss = criterion(output, target)optimizer.zero_grad()loss.backward()optimizer.step()train_loss.append(loss.item())print(f"Epoch [{epoch}/{num_epochs-1}]---train loss{np.mean(train_loss):>.5f}")scheduler.step()if epoch % 10 == 0:validation_loss, validation_acc = validation(model, val_loader, criterion)train_loss_epoch10.append(np.mean(train_loss))val_loss_epoch10.append(validation_loss)val_acc_epoch10.append(validation_acc)train_process["train_loss"] = train_loss_epoch10train_process["val_loss"] = val_loss_epoch10train_process["val_acc"] = val_acc_epoch10return train_processdef validation(model, val_loader, criterion):model.eval()hs = model.init_hidden(batch_size)val_loss = []val_correct = 0with torch.no_grad():for data, target in val_loader:data = data.to(device)target = target.to(device)outs,hs = model(data,hs)hs = tuple([h.data for h in hs])loss = criterion(outs, target)preds = torch.argmax(outs, dim=1)val_loss.append(loss.item())val_correct += torch.sum(preds==target)print(f"--------------------------------validation loss is:{np.mean(val_loss):>.5f}, acc is:{100*val_correct/len(val_loader.dataset):>.2f}%")return np.mean(val_loss), val_correct/len(val_loader.dataset)

train_process = train(model, train_loader, criterion, optimizer,exp_lr_scheduler, num_epochs=100)

Epoch [0/99]---train loss 1.59277
--------------------------------validation loss is: 1.56931, acc is: 34.67%
Epoch [1/99]---train loss 1.58079
Epoch [2/99]---train loss 1.47616
Epoch [3/99]---train loss 1.16114
Epoch [4/99]---train loss 1.00649
Epoch [5/99]---train loss 0.79341
Epoch [6/99]---train loss 0.61482
Epoch [7/99]---train loss 0.37631
Epoch [8/99]---train loss 0.22457
Epoch [9/99]---train loss 0.15812
Epoch [10/99]---train loss 0.10514
--------------------------------validation loss is: 0.46727, acc is: 86.64%
Epoch [11/99]---train loss 0.09292
Epoch [12/99]---train loss 0.08403
Epoch [13/99]---train loss 0.07773
Epoch [14/99]---train loss 0.07114
Epoch [15/99]---train loss 0.07018
Epoch [16/99]---train loss 0.06568
Epoch [17/99]---train loss 0.06244
Epoch [18/99]---train loss 0.06041
Epoch [19/99]---train loss 0.05722
Epoch [20/99]---train loss 0.05333
--------------------------------validation loss is: 0.55406, acc is: 87.14%
Epoch [21/99]---train loss 0.04836
Epoch [22/99]---train loss 0.04972
Epoch [23/99]---train loss 0.04870
Epoch [24/99]---train loss 0.05047
Epoch [25/99]---train loss 0.04990
Epoch [26/99]---train loss 0.05206
Epoch [27/99]---train loss 0.04928
Epoch [28/99]---train loss 0.04778
Epoch [29/99]---train loss 0.04771
Epoch [30/99]---train loss 0.04868
--------------------------------validation loss is: 0.57627, acc is: 87.22%
Epoch [31/99]---train loss 0.04743
Epoch [32/99]---train loss 0.04727
Epoch [33/99]---train loss 0.04861
Epoch [34/99]---train loss 0.04523
Epoch [35/99]---train loss 0.04876
Epoch [36/99]---train loss 0.04664
Epoch [37/99]---train loss 0.04642
Epoch [38/99]---train loss 0.04934
Epoch [39/99]---train loss 0.04657
Epoch [40/99]---train loss 0.04912
--------------------------------validation loss is: 0.57772, acc is: 87.37%
Epoch [41/99]---train loss 0.05022
Epoch [42/99]---train loss 0.04639
Epoch [43/99]---train loss 0.04854
Epoch [44/99]---train loss 0.04836
Epoch [45/99]---train loss 0.04772
Epoch [46/99]---train loss 0.04839
Epoch [47/99]---train loss 0.04706
Epoch [48/99]---train loss 0.04579
Epoch [49/99]---train loss 0.04788
Epoch [50/99]---train loss 0.04893
--------------------------------validation loss is: 0.57018, acc is: 87.59%
Epoch [51/99]---train loss 0.04846
Epoch [52/99]---train loss 0.04869
Epoch [53/99]---train loss 0.04680
Epoch [54/99]---train loss 0.04806
Epoch [55/99]---train loss 0.04735
Epoch [56/99]---train loss 0.04469
Epoch [57/99]---train loss 0.04781
Epoch [58/99]---train loss 0.04836
Epoch [59/99]---train loss 0.04770
Epoch [60/99]---train loss 0.04822
--------------------------------validation loss is: 0.58553, acc is: 87.24%
Epoch [61/99]---train loss 0.04633
Epoch [62/99]---train loss 0.05106
Epoch [63/99]---train loss 0.04976
Epoch [64/99]---train loss 0.04692
Epoch [65/99]---train loss 0.04643
Epoch [66/99]---train loss 0.05045
Epoch [67/99]---train loss 0.04795
Epoch [68/99]---train loss 0.04573
Epoch [69/99]---train loss 0.05088
Epoch [70/99]---train loss 0.04687
--------------------------------validation loss is: 0.57736, acc is: 87.59%
Epoch [71/99]---train loss 0.04938
Epoch [72/99]---train loss 0.05052
Epoch [73/99]---train loss 0.04557
Epoch [74/99]---train loss 0.04864
Epoch [75/99]---train loss 0.05058
Epoch [76/99]---train loss 0.04869
Epoch [77/99]---train loss 0.04648
Epoch [78/99]---train loss 0.04884
Epoch [79/99]---train loss 0.04855
Epoch [80/99]---train loss 0.04667
--------------------------------validation loss is: 0.58926, acc is: 87.09%
Epoch [81/99]---train loss 0.04856
Epoch [82/99]---train loss 0.04642
Epoch [83/99]---train loss 0.04517
Epoch [84/99]---train loss 0.04651
Epoch [85/99]---train loss 0.04770
Epoch [86/99]---train loss 0.04673
Epoch [87/99]---train loss 0.04621
Epoch [88/99]---train loss 0.04786
Epoch [89/99]---train loss 0.04591
Epoch [90/99]---train loss 0.04608
--------------------------------validation loss is: 0.57623, acc is: 87.29%
Epoch [91/99]---train loss 0.04993
Epoch [92/99]---train loss 0.04699
Epoch [93/99]---train loss 0.04799
Epoch [94/99]---train loss 0.04821
Epoch [95/99]---train loss 0.04724
Epoch [96/99]---train loss 0.04790
Epoch [97/99]---train loss 0.04842
Epoch [98/99]---train loss 0.04961
Epoch [99/99]---train loss 0.04863

训练损失可视化

import matplotlib.pyplot as plt
plt.figure(figsize=(16,4))
plt.subplot(1,2,1)
plt.title("Loss")
plt.ylabel("loss")
plt.plot(train_process["train_loss"],label="train-loss")
plt.plot(train_process["val_loss"],label="val-loss")
plt.legend()plt.subplot(1,2,2)
plt.legend(labels=[""])
plt.ylabel("accuracy")
plt.title("Validation Accuracy")
plt.plot(train_process["val_acc"])
plt.show()

​

预测

from string import punctuation
texts = ["i wont let me child cry it out because i feel that loving her and lily when she was little was going to be opportunities that only lasted for those short few months","i feel there are dangerous games or activities","i feel dirty and ashamed for saying that"]
def converts(text):text = remove_hyperlinks(text)text = remove_currencies(text)text = remove_number(text)new_text = remove_punctuation(text)test_text_ints = [word2int[word.lower()]for word in new_text.split() if word in word2int.keys()]return test_text_intsdef predict(model):test_text_ints = [converts(text) for text in texts]new_test_text_ints = reset_text(test_text_ints, seq_len=22)text_tensor = torch.from_numpy(new_test_text_ints)batch_size = text_tensor.size(0)hs = model.init_hidden(batch_size)text_tensor = text_tensor.to(device)outs, hs = model(text_tensor,hs)preds = torch.argmax(outs, dim=1)for i in range(len(texts)):print(texts[i])print("   预测结果是: ", category[int(preds[i])])predict(model)

i wont let me child cry it out because i feel that loving her and lily when she was little was going to be opportunities that only lasted for those short few months预测结果是:  love
i feel there are dangerous games or activities预测结果是:  anger
i feel dirty and ashamed for saying that预测结果是:  sadness

lstm实战，nlp情感分析（Kaggle）相关推荐

  1. RNN、LSTM、GRU实战——imdb情感分析

    Day40 RNN.LSTM.GRU实战--imdb情感分析 RNN Cell模式 Layer模式 LSTM Cell模式 Layer模式 GRU Cell模式 Layer模式 RNN.LSTM.GR ...

  2. python商品评论数据采集与分析可视化系统 Flask框架 requests爬虫 NLP情感分析 毕业设计 源码

    一.项目介绍 python商品评论数据采集与分析可视化系统 Flask框架.MySQL数据库. requests爬虫.可抓取指定商品评论.Echarts可视化.评论多维度分析.NLP情感分析.LDA主 ...

  3. 用户评论想表达什么？NLP情感分析技术告诉你答案

    情感分析是自然语言处理中常见的场景,比如淘宝商品评价,饿了么外卖评价等,对于指导产品更新迭代具有关键性作用.通过情感分析,可以挖掘产品在各个维度的优劣,从而明确如何改进产品. 比如对外卖评价,可以分析 ...

  4. LSTM+word2vec电影情感分析

    LSTM+word2vec电影情感分析 import numpy as np import matplotlib.pyplot as plt import pandas as pd from skle ...

  5. 自然语言处理(NLP)之使用LSTM进行文本情感分析

    情感分析简介 文本情感分析(Sentiment Analysis)是自然语言处理(NLP)方法中常见的应用,也是一个有趣的基本任务,尤其是以提炼文本情绪内容为目的的分类.它是对带有情感色彩的主观性文本 ...

  6. 【深度学习项目五】：利用LSTM网络进行情感分析（NLP）

    相关文章: [深度学习项目一]全连接神经网络实现mnist数字识别](https://blog.csdn.net/sinat_39620217/article/details/116749255?sp ...

  7. 基于LSTM分类文本情感分析

    背景介绍 文本情感分析作为NLP的常见任务,具有很高的实际应用价值.本文将采用LSTM模型,训练一个能够识别文本postive, neutral, negative三种情感的分类器. 本文的目的是快速 ...

  8. lstm训练情感分析的优点_使用LSTM进行文本情感分析

    文本情感分析(Sentiment Analysis)是自然语言处理(NLP)方法中常见的应用,也是一个有趣的基本任务,尤其是以提炼文本情绪内容为目的的分类.它是对带有情感色彩的主观性文本进行分析.处理 ...

  9. lstm训练情感分析的优点_NLP入门（十）使用LSTM进行文本情感分析

    情感分析简介 文本情感分析(Sentiment Analysis)是自然语言处理(NLP)方法中常见的应用,也是一个有趣的基本任务,尤其是以提炼文本情绪内容为目的的分类.它是对带有情感色彩的主观性文本 ...

  10. [深度学习-实战篇]情感分析之卷积神经网络-TextCNN,包含代码

    0. 前言 在"卷积神经网络"中我们探究了如何使用二维卷积神经网络来处理二维图像数据.在之前的语言模型和文本分类任务中,我们将文本数据看作是只有一个维度的时间序列,并很自然地使用循 ...

最新文章

  1. python 获取当前ip
  2. SylixOS内核打印调试方法
  3. .NET Entity Framework入门简介及简单操作
  4. 秒懂词向量Word2vec的本质
  5. msp430入门学习31
  6. java 图片分段上传_java文件分片上传,断点续传
  7. linux ksh 历史命令,防止Linux中的bash(和ksh)历史记录更改
  8. C++ 11 深度学习（三）范围for、new内存动态分配、nullptr
  9. nginx tcp转发_Nginx学习（九）：负载均衡服务
  10. 【ZOJ - 1163】The Staircases（dp）
  11. mysql基本介绍和优化技巧
  12. 澎思科技新出行人再识别（ReID）算法，刷新三大数据集最高记录
  13. Directx11教程(65) 渲染到纹理
  14. 启动hive报错_远程服务器模式Hive的搭建过程详解
  15. Git子模块头#39；引用不是树#39；错误
  16. 【windows核心编程】IO完成端口(IOCP)复制文件小例
  17. 基于react-app搭建react-router+redux项目
  18. 数据挖掘中最易犯的10个错误，请绕行！
  19. HTML中让整个效果居中,仅使用CSS做到完全居中的超级攻略
  20. jiaba库之关键词提取（增量更新自定义语料）

热门文章

  1. 【我的C/C++语言学习进阶之旅】NDK开发之解决错误：signal 5 (SIGTRAP), code 1 (TRAP_BRKPT), fault addr 0xXXX
  2. oracle服务器配置端口,服务端配置一个多端口号（如1521/1526）监听器
  3. PIC单片机学习笔记——PIC16F(L)1526/7
  4. 嵌入式系统开发笔记25：win10防火墙打开特定端口
  5. 【今日CV 计算机视觉论文速览 第147期】Tue, 23 Jul 2019
  6. linux查看UID命令,Linux下查看UID
  7. Docker基础（centos安装、阿里云镜像加速器配置、常用命令、数据卷、DockerFile、镜像推送阿里云等）
  8. puppet插件fact和hiera（puppet自动化系列3）
  9. 汽车电瓶电压12V验证
  10. nginx ngx_event_t结构体详解