在好例子网,分享、交流、成长!
您当前所在位置:首页Others 开发实例Clojure → 斯坦福CS224n_自然语言处理与深度学习_笔记_hankcs

斯坦福CS224n_自然语言处理与深度学习_笔记_hankcs

Clojure

下载此实例
  • 开发语言:Others
  • 实例大小:37.23M
  • 下载次数:8
  • 浏览次数:28
  • 发布时间:2022-07-23
  • 实例类别:Clojure
  • 发 布 人:huancai19780315
  • 文件格式:.pdf
  • 所需积分:2

实例介绍

【实例简介】斯坦福CS224n_自然语言处理与深度学习_笔记_hankcs

【实例截图】

【核心代码】


目 录
笔记 1 自然语言处理与深度学习简介..........................................................................................13
1.1 新旧 CS224 对比................................................................................................................13
1.2 CS224 资料.......................................................................................................................14
1.3 什么是自然语言处理........................................................................................................15
1.4 自然语言处理应用............................................................................................................16
1.5 人类语言的特殊之处........................................................................................................17
1.6 什么是深度学习................................................................................................................17
1.7 “深度学习”的历史............................................................................................................. 19
1.8 为什么需要研究深度学习................................................................................................19
1.9 语音识别中的深度学习....................................................................................................19
1.20 计算机视觉中的深度学习............................................................................................. 20
1.21 课程相关..........................................................................................................................21
1.22 为什么 NLP 难................................................................................................................. 21
1.23 Deep NLP = Deep Learning NLP.................................................................................. 21
1.24 word vector.................................................................................................................... 22
1.25 NLP 表示层次:形态级别............................................................................................ 22
1.26 NLP 工具:句法分析.................................................................................................... 22
1.27 NLP 语义层面的表示.................................................................................................... 23
1.28 情感分析..........................................................................................................................23
1.29 QA...................................................................................................................................24
1.30 客服系统..........................................................................................................................25
1.31 机器翻译..........................................................................................................................26
1.32 结论:所有层级的表示都是向量................................................................................. 27
研究热点 1 一个简单但很难超越的 Sentence Embedding 基线方法.........................................29
1.1 句子 Embedding 动机....................................................................................................... 29
1.2 已有方法............................................................................................................................30
1.3 新方法................................................................................................................................31
1.4 概率论解释........................................................................................................................32
1.5 效果....................................................................................................................................32
笔记 2 词的向量表示:word2vec..................................................................................................34
2.1 如何表示一个词语的意思................................................................................................34
2.2 计算机如何处理词语的意思........................................................................................... 34
2.3 discrete representation 的问题.......................................................................................34
2.4 从 symbolic representations 到 distributed representations...........................................34
2.5 Distributional similarity based representations.............................................................. 35
2.6 通过向量定义词语的含义................................................................................................35
2.7 学习神经网络 word embeddings 的基本思路................................................................35
2.8 直接学习低维词向量........................................................................................................36
2.9 word2vec 的主要思路.....................................................................................................36
2.10 Skip-gram 预测.............................................................................................................. 36
2.11 word2vec 细节...............................................................................................................37
2.12 目标函数细节..................................................................................................................372.13 Word2Vec 细节..............................................................................................................37
2.14 点积..................................................................................................................................37
2.15 Softmax function:从实数空间到概率分布的标准映射方法................................... 38
2.16 Skipgram.........................................................................................................................38
2.17 训练模型:计算参数向量的梯度................................................................................. 40
2.18 损失/目标函数................................................................................................................ 43
2.19 梯度下降、SGD...............................................................................................................43
研究热点 2 词语义项的线性代数结构与词义消歧..................................................................... 45
2.1 复原....................................................................................................................................46
2.2 复原结果............................................................................................................................46
2.3 量化评测............................................................................................................................46
2.4 总结....................................................................................................................................47
笔记 3 高级词向量表示..................................................................................................................48
3.1 复习:word2vec 的主要思路...........................................................................................48
3.2 SGD 与词向量..................................................................................................................48
3.3 近似:负采样....................................................................................................................49
3.4 negative sampling 和 skip-gram...................................................................................... 49
3.5 其他方法............................................................................................................................50
3.6 基于窗口的共现矩阵........................................................................................................50
3.7 朴素共现向量的问题........................................................................................................51
3.8 解决办法:低维向量........................................................................................................51
3.9 改进....................................................................................................................................52
3.10 效果..................................................................................................................................52
3.11 SVD 的问题.................................................................................................................... 54
3.12 Count based vs direct prediction...................................................................................54
3.13 综合两者优势:GloVe....................................................................................................54
3.14 评测方法..........................................................................................................................55
3.15 Intrinsic word vector evaluation.................................................................................... 55
3.16 结果对比..........................................................................................................................57
3.17 调参..................................................................................................................................57
3.18 另一个数据集..................................................................................................................59
3.19 Extrinsic word vector evaluation................................................................................... 60
笔记 4 Word Window 分类与神经网络.......................................................................................62
4.1 分类问题............................................................................................................................62
4.2 softmax 详细....................................................................................................................62
4.3 softmax 与交叉熵误差....................................................................................................62
4.4 优化....................................................................................................................................63
4.5 re-training 词向量失去泛化效果................................................................................... 64
4.6 Window classification...................................................................................................... 65
4.6 最简单的分类器:softmax.............................................................................................. 65
4.7 softmax(等价于逻辑斯谛回归)效果有限................................................................ 66
4.8 使用神经网络....................................................................................................................67
4.9 从逻辑斯谛回归到神经网络........................................................................................... 67
4.10 为什么需要非线性..........................................................................................................704.11 前向传播网络..................................................................................................................71
4.12 间隔最大化目标函数......................................................................................................71
4.13 反向传播训练..................................................................................................................72
研究热点 3 高效文本分类的锦囊妙计..........................................................................................74
3.1 Facebook 的 fastText .......................................................................................................74
3.2 词袋模型............................................................................................................................74
3.3 简单的线性模型................................................................................................................75
3.4 训练....................................................................................................................................75
3.5 Hierarchical softmax........................................................................................................ 76
3.6 效果与速度........................................................................................................................76
3.7 总结....................................................................................................................................77
笔记 5 反向传播与项目指导..........................................................................................................78
5.1 任意层的通用公式............................................................................................................78
5.2 反向传播的电路解释........................................................................................................78
5.3 第三种理解:流程图........................................................................................................82
5.4 第四种解释:实际神经网络中的误差信号................................................................... 86
5.5 课程项目............................................................................................................................87
研究热点 4 词嵌入对传统方法的启发..........................................................................................88
4.1 词语表示方法....................................................................................................................88
4.2 Skip-Gram 中的超参数....................................................................................................89
4.3 对 PMI 的启发................................................................................................................... 89
4.4 可调超参数一览表............................................................................................................90
4.5 调参结果............................................................................................................................90
4.6 结论....................................................................................................................................91
笔记 6 句法分析..............................................................................................................................92
6.1 语言学的两种观点............................................................................................................92
6.2 歧义....................................................................................................................................92
6.3 依附歧义............................................................................................................................93
6.4 标注数据集的崛起:Universal Dependencies 6.5 treebanks......................................... 93
6.6 依存文法与依存结构........................................................................................................94
6.7 起源....................................................................................................................................95
6.8 一些细节............................................................................................................................96
6.9 句法分析可用的特征........................................................................................................96
6.10 依存句法分析..................................................................................................................96
6.11 Arc-standard transition.................................................................................................. 96
6.12 MaltParser......................................................................................................................97
6.13 传统特征表示..................................................................................................................97
6.14 效果评估..........................................................................................................................97
6.15 投射性..............................................................................................................................98
6.16 为什么需要神经网络句法分析器................................................................................. 98
6.17 神经网络依存句法分析器............................................................................................. 99
6.18 为何需要非线性............................................................................................................100
6.19 未来工作........................................................................................................................101
Assignment 1................................................................................................................................... 1021.1 Softmax.......................................................................................................................... 102
1.1.1 softmax 常数不变性.......................................................................................... 102
1.1.2 Python 实现........................................................................................................102
1.2 神经网络基础..................................................................................................................103
1.2.1 sigmoid 梯度.......................................................................................................103
1.2.2 交叉熵损失函数的梯度...................................................................................... 104
1.2.3 推导三层网络的梯度.......................................................................................... 105
1.2.4 参数数量...............................................................................................................106
1.2.5 实现 sigmoid 函数................................................................................................106
1.2.6 实现梯度检查.......................................................................................................107
1.2.7 实现前向传播和反向传播.................................................................................. 108
研究热点 5 图像对话....................................................................................................................110
5.1 相关工作..........................................................................................................................110
5.2 图像视频自动标题..........................................................................................................110
5.3 图像语义对齐..................................................................................................................111
5.4 图像 QA............................................................................................................................111
5.5 贡献..................................................................................................................................112
5.6 技术细节..........................................................................................................................112
5.7 数据集..............................................................................................................................113
5.8 结果..................................................................................................................................113
笔记 7 TensorFlow 入门.............................................................................................................. 115
7.1 深度学习框架简介..........................................................................................................115
7.2 TF 是什么.......................................................................................................................115
7.3 图计算编程模型..............................................................................................................116
7.4 图在哪里..........................................................................................................................117
7.5 如何运行..........................................................................................................................118
7.6 训练..................................................................................................................................118
7.7 如何计算梯度..................................................................................................................118
7.8 变量共享..........................................................................................................................119
7.9 总结..................................................................................................................................119
7.10 现场演示........................................................................................................................120
研究热点 6 基于转移的神经网络句法分析的结构化训练....................................................... 121
6.1 什么是 SyntaxNet............................................................................................................ 121
6.2 项贡献..............................................................................................................................121
6.3 Tri-Training:利用未标注数据..................................................................................... 121
6.4 模型改进..........................................................................................................................122
6.5 结构化感知机训练与柱搜索......................................................................................... 122
6.6 结论...................................................................................................................................123
笔记 8 RNN 和语言模型............................................................................................................. 125
8.1 语言模型..........................................................................................................................125
8.2 传统语言模型..................................................................................................................125
8.3 Recurrent Neural Networks........................................................................................... 126
8.4 损失函数..........................................................................................................................128
8.5 训练 RNN 很难................................................................................................................ 1288.6 梯度消失实例..................................................................................................................130
8.7 防止梯度爆炸..................................................................................................................132
8.8 减缓梯度消失..................................................................................................................132
8.9 困惑度结果......................................................................................................................133
8.10 问题:softmax 太大了................................................................................................. 133
8.11 最后的实现技巧............................................................................................................134
8.12 序列模型的应用............................................................................................................134
8.13 Bidirectional RNNs....................................................................................................... 135
8.14 Deep Bidirectional RNNs..............................................................................................135
8.15 评测................................................................................................................................136
8.16 应用:RNN 机器翻译模型........................................................................................... 136
8.17 回顾................................................................................................................................138
研究热点 7 迈向更好的语言模型................................................................................................139
7.1 更好的输入......................................................................................................................139
7.2 更好的正则化和预处理..................................................................................................140
7.3 更好的模型?..................................................................................................................140
笔记 9 机器翻译和高级 LSTM 及 GRU........................................................................................ 142
9.1 机器翻译..........................................................................................................................142
9.2 传统统计机器翻译系统..................................................................................................142
9.3 第一步:对齐..................................................................................................................143
9.4 对齐之后..........................................................................................................................145
9.5 解码:在海量假设中搜索最佳选择............................................................................. 146
9.6 传统机器翻译..................................................................................................................146
9.7 深度学习来救场..............................................................................................................146
9.8 主要改进:更好的单元..................................................................................................150
9.9 LSTM(
Long-Short-Term-Memories)......................................................................... 151
9.10 一些可视化....................................................................................................................153
9.11 LSTM 很潮....................................................................................................................154
9.12 深度 LSTM 用于机器翻译.............................................................................................156
9.13 进一步改进:更多门!............................................................................................... 157
9.14 RNN 的最新改进......................................................................................................... 157
9.15 softmax 的问题:无法出新词................................................................................... 157
9.16 用指针来解决问题........................................................................................................157
9.17 总结................................................................................................................................159
Assignment 2................................................................................................................................... 160
2.1 Tensorflow Softmax........................................................................................................160
2.1.1 softmax................................................................................................................160
2.1.2 交叉熵...................................................................................................................161
2.1.3 Placeholders & Feed Dictionaries.......................................................................161
2.1.4 Softmax & CE Loss...............................................................................................161
2.1.5 Training Optimizer...............................................................................................162
研究热点 8 谷歌的多语种神经网络翻译系统........................................................................... 164
8.1 双语 NMT.........................................................................................................................164
8.2 土办法..............................................................................................................................1648.3 Google 的多语种 NMT 系统.........................................................................................165
8.4 架构..................................................................................................................................166
8.5 效果..................................................................................................................................167
8.6 Zero-Shot Translation.....................................................................................................167
笔记 10 NMT 与 Attention.......................................................................................................... 168
10.1 机器翻译........................................................................................................................168
10.2 机器翻译的需求............................................................................................................168
10.3 什么是 NMT...................................................................................................................168
10.4 架构................................................................................................................................168
10.5 NMT:青铜时代.............................................................................................................168
10.6 现代 NMT 模型..............................................................................................................170
10.7 RNN Encoder................................................................................................................ 171
10.8 Decoder:递归语言模型............................................................................................172
10.9 MT 的发展................................................................................................................... 172
10.10 NMT 的四大优势.......................................................................................................172
10.11 统计/神经网络机器翻译............................................................................................173
10.12 NMT 主要由工业界促进.......................................................................................... 173
10.13 介绍 Attention............................................................................................................. 174
10.14 Attention 机制........................................................................................................... 174
10.15 词语对齐......................................................................................................................174
10.16 同时学习翻译和对齐................................................................................................. 175
10.17 打分..............................................................................................................................176
10.18 更多 attention!覆盖范围......................................................................................... 180
10.19 Doubly attention........................................................................................................ 180
10.20 用旧模型的语言学思想拓展 attention..................................................................... 181
10.21 decoder...................................................................................................................... 181
10.22 Ancestral sampling.....................................................................................................182
10.23 Greedy Search............................................................................................................182
10.24 Beam search...............................................................................................................183
10.25 效果对比......................................................................................................................183
研究热点 9 读唇术........................................................................................................................185
9.1 唇语翻译..........................................................................................................................185
9.2 架构..................................................................................................................................186
9.3 视觉..................................................................................................................................187
9.4 听觉..................................................................................................................................187
9.5 Attention 与 Spell...........................................................................................................188
9.6 Curriculum Learning.......................................................................................................188
9.7 Scheduled Sampling.......................................................................................................189
9.8 数据集..............................................................................................................................189
9.9 结果..................................................................................................................................190
笔记 11 GRU 和 NMT 的进一步话题......................................................................................... 191
11.1 深入 GRU .......................................................................................................................191
11.2 Update Gate................................................................................................................. 191
11.3 Reset Gate.................................................................................................................... 19111.4 GRU 寄存器................................................................................................................. 192
11.5 GRU 和 LSTM 对比...................................................................................................... 192
11.6 深入 LSTM......................................................................................................................193
11.7 训练技巧........................................................................................................................198
11.8 Ensemble......................................................................................................................198
11.9 MT 评测....................................................................................................................... 198
11.10 BLEU........................................................................................................................... 199
11.11 Brevity Penalty........................................................................................................... 199
11.12 Multiple Reference Translations................................................................................ 200
11.13 解决大词表问题..........................................................................................................202
11.14 Large-vocab NMT....................................................................................................... 203
11.15 训练..............................................................................................................................203
11.16 测试..............................................................................................................................204
11.17 更多技巧......................................................................................................................204
11.18 Byte Pair Encoding..................................................................................................... 204
11.19 其他..............................................................................................................................205
笔记 12 语音识别的 end-to-end 模型......................................................................................... 208
12.1 Automatic Speech Recognition(
ASR)..................................................................... 208
12.2 传统 ASR........................................................................................................................ 208
12.3 近现代 ASR.................................................................................................................... 208
12.4 end-to-end ASR............................................................................................................ 209
12.5 Connectionist Temporal Classification.........................................................................209
12.6 一些效果........................................................................................................................210
12.7 sequence to sequence speech recognition with attention.........................................211
12.8 Listen Attend and Spell................................................................................................ 212
12.9 效果................................................................................................................................214
12.10 LAS 的限制.................................................................................................................214
12.11 在线 seq2seq 模型...................................................................................................... 215
12.12 Neural Transducer......................................................................................................215
12.13 结果..............................................................................................................................217
12.14 Encoder 中的卷积..................................................................................................... 218
12.15 目标颗粒度..................................................................................................................219
12.16 效果..............................................................................................................................220
12.17 模型缺点......................................................................................................................221
12.18 解决办法......................................................................................................................221
12.19 另一个缺点..................................................................................................................222
12.20 Better Language Model Blending.............................................................................. 223
12.21 Better Sequence Training...........................................................................................223
12.22 机会..............................................................................................................................223
12.23 多音源..........................................................................................................................224
12.24 "同声传译"....................................................................................................................224
Assignment 3................................................................................................................................... 226
3.1 命名实体识别初步..........................................................................................................226
3.2 A window into NER.........................................................................................................2263.2.1 概念.......................................................................................................................227
3.2.2 维度和复杂度.......................................................................................................229
3.2.3 实现基线模型.......................................................................................................229
3.2.4 分析结果...............................................................................................................230
研究热点 10 Character-Aware 神经网络语言模型...................................................................232
10.1 动机................................................................................................................................232
10.2 架构................................................................................................................................232
10.3 卷积层............................................................................................................................233
10.4 Highway Network.........................................................................................................233
10.5 LSTM.............................................................................................................................234
10.6 量化结果........................................................................................................................234
10.7 直观效果........................................................................................................................235
10.8 结论................................................................................................................................236
10.9 实现................................................................................................................................236
CS224n 笔记 13 卷积神经网络.................................................................................................... 237
13.1 从 RNN 到 CNN..............................................................................................................237
13.2 什么是卷积....................................................................................................................237
13.3 单层 CNN....................................................................................................................... 238
13.4 池化................................................................................................................................239
13.5 分类................................................................................................................................239
13.6 图示................................................................................................................................239
13.7 dropout.........................................................................................................................240
13.8 试验结果........................................................................................................................240
13.9 CNN 花样..................................................................................................................... 241
13.10 CNN 应用:机器翻译............................................................................................... 242
13.11 模型比较......................................................................................................................242
13.12 Quasi-RNN..................................................................................................................243
研究热点 11 深度强化学习用于对话生成................................................................................. 245
11.1 seq2seq 的缺陷 .......................................................................................................... 245
11.2 如何定义好的回复........................................................................................................246
11.3 强化学习........................................................................................................................246
11.4 量化结果........................................................................................................................247
11.5 直观效果........................................................................................................................247
11.6 结论................................................................................................................................248
笔记 14 Tree RNN 与短语句法分析........................................................................................... 249
14.1 语言模型光谱 ...............................................................................................................249
14.2 语言的语义解释——并不只是词向量....................................................................... 249
14.3 语义合成性....................................................................................................................250
14.4 语言能力........................................................................................................................251
14.5 语言是递归的吗............................................................................................................251
14.6 在词向量空间模型上表示语义................................................................................... 252
14.7 如何将短语映射到向量空间....................................................................................... 252
14.8 短语结构分析:目的....................................................................................................253
14.9 Recursive vs. recurrent neural networks..................................................................... 25414.10 从 RNNs 到 CNNs.........................................................................................................255
14.11 Recursive Neural Networks 用于结构化预测.......................................................... 255
14.12 最简单的 Recursive Neural Network.......................................................................... 256
14.13 用 RNN 分析句子........................................................................................................ 256
14.14 最大间隔......................................................................................................................258
14.15 结构上的反向传播..................................................................................................... 258
14.16 简单 RNN 的缺点........................................................................................................ 259
研究热点 12 神经网络自动代码摘要......................................................................................... 261
12.1 任务与数据集 ...............................................................................................................261
12.2 子任务 ..........................................................................................................................261
12.3 网络架构 .......................................................................................................................262
12.4 结果................................................................................................................................263
12.5 量化评测........................................................................................................................263
12.6 直观效果........................................................................................................................263
笔记 15 指代消解..........................................................................................................................265
15.1 什么是指代消解 ...........................................................................................................265
15.2 应用................................................................................................................................265
15.3 指代消解评测................................................................................................................266
15.4 指代的类型....................................................................................................................266
15.5 不是所有 NP 都在指代.................................................................................................266
15.6 Coreference, anaphors, cataphors...............................................................................267
15.7 共指与回指....................................................................................................................267
15.7 传统代词消解方法:Hobbs’naive algorithm........................................................... 268
15.8 基于知识库的代词消解............................................................................................... 268
15.9 几种指代消解模型........................................................................................................268
15.10 监督 Mention-Pair Model............................................................................................268
15.11 指代消解可用特征..................................................................................................... 269
15.12 神经网络指代消解模型............................................................................................. 269
研究热点 13 学习代码的语义......................................................................................................270
13.1 表示代码 .......................................................................................................................270
13.2 编码解码状态................................................................................................................271
13.3 目标函数........................................................................................................................272
13.4 利用 RecursiveNN 来生成程序 embeddings................................................................272
13.5 总结................................................................................................................................272
13.6 未来工作........................................................................................................................273
笔记 16 DMN 与问答系统.......................................................................................................... 274
16.1 是否所有 NLP 任务都可视作 QA?.............................................................................274
16.2 前无古人........................................................................................................................274
16.3 全才难得........................................................................................................................274
16.4 Dynamic Memory Networks........................................................................................275
16.5 回答难题........................................................................................................................275
16.6 Dynamic Memory Networks........................................................................................275
16.7 The Modules: Input......................................................................................................276
16.8 The Modules: Question................................................................................................27716.9 The Modules: Episodic Memory..................................................................................277
16.10 The Modules: Answer................................................................................................ 279
16.11 相关工作......................................................................................................................279
16.12 与 MemNets 比较....................................................................................................... 279
研究热点 14 自动组合神经网络做问答系统............................................................................. 281
14.1 四个 Jointly 训练的组件............................................................................................... 281
14.2 模型:在两个分布上构建........................................................................................... 282
14.3 Layout Model............................................................................................................... 282
14.4 Layout Scoring Model.................................................................................................. 284
14.5 Execution Model.......................................................................................................... 284
14.5 Module: lookup............................................................................................................285
14.6 Module: relate..............................................................................................................285
14.7 Module: find.................................................................................................................286
14.8 Module: and.................................................................................................................286
14.9 训练 Execution Model................................................................................................... 287
14.10 结果..............................................................................................................................287
14.11 VQA............................................................................................................................ 287
14.12 GeoQA........................................................................................................................ 288
笔记 17 NLP 存在的问题与未来的架构....................................................................................... 289
17.1 新时代人们正在“解决”语言 .................................................................................. 289
17.2 旧时代的热血................................................................................................................289
17.3 基础 NLP:在进步........................................................................................................ 290
17.4 我们还需要什么............................................................................................................291
17.5 Recursive Neural Networks 用于意识形态检测........................................................ 292
17.6 TreeRNN....................................................................................................................... 293
17.7 recurrent NN 训练快................................................................................................... 293
17.8 TreeRNN 结构取决于输入.......................................................................................... 294
17.9 The Shift-reduce Parser-Interpreter NN (SPINN).........................................................294
17.10 binary trees = transition sequences.......................................................................... 294
17.11 架构..............................................................................................................................295
研究热点 15 Neural Turing Machines ........................................................................................ 296
15.1 问题 ...............................................................................................................................296
15.2 记忆是解决方案吗........................................................................................................297
15.3 Neural Turing Machines...............................................................................................297
15.4 如何读写........................................................................................................................300
15.5 读内存............................................................................................................................300
15.6 写内存............................................................................................................................301
15.7 attention 更新..............................................................................................................301
15.8 第一步............................................................................................................................301
15.9 第二步............................................................................................................................302
15.10 第三步..........................................................................................................................302
15.11 效果..............................................................................................................................303
15.12 References..................................................................................................................303
笔记 18 挑战深度学习与自然语言处理的极限......................................................................... 30418.1 障碍 1:通用架构 ........................................................................................................304
18.2 障碍 2:联合多任务学习............................................................................................ 305
18.2.1 解决方案..............................................................................................................305
18.2.2 模型细节..............................................................................................................306
18.2.3 依存句法分析..................................................................................................... 307
18.2.4 语义联系..............................................................................................................307
18.2.5 训练......................................................................................................................308
18.2.6 结果......................................................................................................................308
18.3 障碍 3:预测从未见过的词语.................................................................................... 309


实例下载地址

斯坦福CS224n_自然语言处理与深度学习_笔记_hankcs

不能下载?内容有错? 点击这里报错 + 投诉 + 提问

好例子网口号:伸出你的我的手 — 分享

网友评论

发表评论

(您的评论需要经过审核才能显示)

查看所有0条评论>>

小贴士

感谢您为本站写下的评论,您的评论对其它用户来说具有重要的参考价值,所以请认真填写。

  • 类似“顶”、“沙发”之类没有营养的文字,对勤劳贡献的楼主来说是令人沮丧的反馈信息。
  • 相信您也不想看到一排文字/表情墙,所以请不要反馈意义不大的重复字符,也请尽量不要纯表情的回复。
  • 提问之前请再仔细看一遍楼主的说明,或许是您遗漏了。
  • 请勿到处挖坑绊人、招贴广告。既占空间让人厌烦,又没人会搭理,于人于己都无利。
;
报警