1
00:00:06,875 --> 00:00:10,825
同学们 欢迎来到自然语言处理在线课程
Dear students  welcome to  Natural Language Processing online course

2
00:00:11,100 --> 00:00:13,100
我是西湖大学的张岳 
I'm Yue Zhang from Westlake University

3
00:00:13,400 --> 00:00:16,250
这节课介绍课程的整体内容
This lesson is an overall introduction to our Course 

4
00:00:19,170 --> 00:00:22,700
这节课是咱们这门课的一个整体介绍
This lesson is an overall introduction to our Course

5
00:00:27,320 --> 00:00:29,890
我们的课程叫做自然语言处理
The name of our course is natural language processing

6
00:00:30,050 --> 00:00:31,490
natural language processing
natural language processing

7
00:00:32,170 --> 00:00:33,050
或者 nlp
or NLP in its abbreviation

8
00:00:34,910 --> 00:00:35,760
自然语言处理
The field of Natural Language Processing

9
00:00:35,760 --> 00:00:37,680
是人工智能的一个重要分支
is an important branch of Artificial Intelligence

10
00:00:38,470 --> 00:00:39,040
他研究
which studies

11
00:00:39,190 --> 00:00:41,600
机器如何自动的去理解和生成
how does machine understand and generate

12
00:00:41,710 --> 00:00:42,480
人类的语言
human language automatically

13
00:00:44,140 --> 00:00:44,600
那么早
So back to the early stage

14
00:00:44,600 --> 00:00:47,570
在计算机科学和人工智能发展的
of Computer Science and Artificial Intelligence

15
00:00:48,060 --> 00:00:48,770
初期阶段
research

16
00:00:49,500 --> 00:00:52,130
自然语言处理就受到了学界的关注
NLP has attracted attention from the Community

17
00:00:53,180 --> 00:00:54,970
比如人们会开始探究
for example, they began to study

18
00:00:55,340 --> 00:00:56,240
如何自动的
how to translate automatically

19
00:00:56,240 --> 00:00:58,570
将一种语言翻译成另外一种语言
from one language to another

20
00:01:00,480 --> 00:01:01,390
那么最近呢
Meanwhile, recently

21
00:01:01,920 --> 00:01:03,630
随着深度学习技术
with the rapid development of deep learning technologies

22
00:01:04,320 --> 00:01:06,270
和预训练技术的快速发展
and pre-training methods

23
00:01:07,000 --> 00:01:09,590
自然语言处理也获得了长足的进步
NLP has gained a lot of success

24
00:01:10,690 --> 00:01:11,720
我们可以看到
We can see products like

25
00:01:12,290 --> 00:01:13,160
智能音箱
Smart Speakers

26
00:01:13,770 --> 00:01:14,600
机器翻译
Machine Translation Systems

27
00:01:15,130 --> 00:01:15,880
自动文摘
Automatic Summarization

28
00:01:16,570 --> 00:01:18,160
作文自动评分等等
Automatic essay scoring

29
00:01:18,450 --> 00:01:20,080
需要自然语言处理的系统
Systems that require NLP technology

30
00:01:20,330 --> 00:01:22,960
在工业界也越来越多的得到了应用
are also being implemented more often in the industry

31
00:01:26,410 --> 00:01:28,880
那么自然语言处理是一个机器学习
NLP is a machine learning (ML) driven

32
00:01:29,050 --> 00:01:30,240
驱动的学科
Subject

33
00:01:35,710 --> 00:01:37,400
在我们这门课程里呢
In our course

34
00:01:37,950 --> 00:01:38,480
我们会
we will

35
00:01:39,030 --> 00:01:41,960
非常综合的介绍自然语言处理的
give a comprehensive introduction to

36
00:01:42,430 --> 00:01:43,880
基本方法和内容
the basics of NLP disciplines and methods

37
00:01:45,350 --> 00:01:46,480
我们会围绕着
We will concentrate on

38
00:01:46,710 --> 00:01:49,280
其中这个关键的机器学习技术
those key machine learning concepts

39
00:01:49,670 --> 00:01:51,040
展开课程的讨论
for our discussion

40
00:01:55,940 --> 00:01:57,070
我们的课程呢
Our course

41
00:01:57,420 --> 00:01:58,190
既可以被
is both suitable for

42
00:01:59,060 --> 00:02:01,270
高年级的本科生选用
senior undergraduate students

43
00:02:02,300 --> 00:02:03,070
也可以为
and also for

44
00:02:03,460 --> 00:02:05,150
准备从事自然语言处理
junior graduate students

45
00:02:05,220 --> 00:02:08,030
或者相关的人工智能学科的研究的
who are carrying on researches

46
00:02:08,580 --> 00:02:09,630
低年级的
in NLP

47
00:02:10,420 --> 00:02:12,750
硕士生博士生所采用
and Artificial Intelligence related fields

48
00:02:14,280 --> 00:02:14,890
另外呢
Besides

49
00:02:15,760 --> 00:02:16,850
ai 工程师
If you are AI engineers

50
00:02:17,320 --> 00:02:19,450
如果想深入了解自然语言处理
That are interested in

51
00:02:19,480 --> 00:02:20,890
背后的技术和方法
the technologies and methods backing NLP applications

52
00:02:21,440 --> 00:02:22,770
能够提出新的模型
and able to design original models

53
00:02:23,520 --> 00:02:24,890
这门课程也会对你
This course is also

54
00:02:25,320 --> 00:02:26,570
起到一定的帮助作用
dedicated to you

55
00:02:31,900 --> 00:02:34,050
我们的课程总结起来呢
To summarize

56
00:02:34,820 --> 00:02:36,250
有六个主要的特点
Our course has 6 major features

57
00:02:38,020 --> 00:02:40,570
首先是我们的讨论会比较深入
first our discussion will be thorough

58
00:02:41,970 --> 00:02:43,840
除了介绍自然语言处理
We will not only talk about

59
00:02:43,890 --> 00:02:45,480
主要研究的任务以外
the well-known tasks of NLP

60
00:02:46,250 --> 00:02:48,880
我们会以机器学习的方法
But we will focus on the ML technologies

61
00:02:49,570 --> 00:02:51,320
和数学背景为中心
and the mathematical principles inside

62
00:02:52,050 --> 00:02:53,080
仔细的阐明
to illustrate

63
00:02:54,010 --> 00:02:56,640
解决各大类任务所需要的
How the fundamental

64
00:02:57,010 --> 00:03:00,200
典型的数据结构和算法的基础
data structures and algorithms work in these tasks

65
00:03:02,900 --> 00:03:06,530
第二个特点是我们的涵盖会比较全面
The second feature is that our coverage would be broad

66
00:03:08,140 --> 00:03:09,170
这门课程呢
This course

67
00:03:09,420 --> 00:03:11,490
将会讨论自然语言处理
will get to all the major

68
00:03:11,820 --> 00:03:13,490
这门学科发展过程中
methods of NLP

69
00:03:14,020 --> 00:03:16,370
所涉及到的所有主要的方法
during its evolvement

70
00:03:17,930 --> 00:03:18,820
那既包括
That include

71
00:03:19,410 --> 00:03:21,660
传统意义上的统计学习方法
The traditional statistical learning

72
00:03:22,370 --> 00:03:24,820
又包括最近的深度学习方法
the recent deep learning methods

73
00:03:26,530 --> 00:03:29,540
那么既包括一些生成式的数学模型
We will look into both generative mathematical models

74
00:03:30,250 --> 00:03:32,860
又包括一些判别式的数学模型
and discriminative mathematical models

75
00:03:36,140 --> 00:03:37,330
第三个特点呢
The third feature is that

76
00:03:37,820 --> 00:03:41,890
是我们的课程尽量的设计的浅显易懂
We want our course to be clear and easy to understand

77
00:03:43,060 --> 00:03:46,250
当我们有比较陡的学习坡度的时候
When we have a relatively steep learning curve

78
00:03:46,580 --> 00:03:49,570
我们尽量给出尽可能多的细节
we will provide more details to it

79
00:03:50,220 --> 00:03:52,490
让学生能够循序渐进
We hope you can learn it step by step

80
00:03:55,790 --> 00:03:58,560
第四个特点是我们课程的组织方式

81
00:04:00,150 --> 00:04:01,240
我们的课程呢
The fourth feature is that our course

82
00:04:01,670 --> 00:04:02,320
整体上
As a whole

83
00:04:02,750 --> 00:04:05,680
是按照整个这门学科发展的历程来
is designed according to the timeline

84
00:04:06,030 --> 00:04:06,680
组织的
of NLP progress

85
00:04:07,510 --> 00:04:08,480
那么模型呢
The models we introduce

86
00:04:09,410 --> 00:04:12,640
也从早期研究的相对较为简单的
would begin with the simple

87
00:04:13,130 --> 00:04:14,000
统计方法
statistical methods

88
00:04:15,130 --> 00:04:18,440
到最近的大规模预训练
to the recent large-scale pre-training models

89
00:04:19,290 --> 00:04:21,120
还有深度引变量模型
and deep latent variable models

90
00:04:23,770 --> 00:04:25,580
所有这些典型的方法呢
All these classical models

91
00:04:25,890 --> 00:04:29,380
都按照他被发明和出现的时间顺序
will be introduced based on the timeline

92
00:04:29,610 --> 00:04:31,300
这个大体框架来组织
of their inventions or first appearances

93
00:04:32,090 --> 00:04:33,580
这个也符合模型
which also corresponds to our principal of

94
00:04:33,610 --> 00:04:36,380
从简单到复杂的一个逻辑顺序
from simple to complex

95
00:04:40,010 --> 00:04:41,670
课程的第五个特点呢
The fifth feature is that

96
00:04:42,780 --> 00:04:45,750
是我们用一个统一的数学符号框架
all the models that we discuss

97
00:04:46,220 --> 00:04:48,030
来涵盖了
are unified

98
00:04:48,600 --> 00:04:50,875
所有的被讨论到的模型
under the universal mathematical symbol framework

99
00:04:52,190 --> 00:04:55,040
那么不管是简单的统计模型
No matter if it's simple statistic models

100
00:04:56,110 --> 00:04:58,720
还是参数量更大的深度学习模型
Or it is deep learning models with a ton of parameters

101
00:04:59,740 --> 00:05:01,730
不管是有监督学习的模型
No matter it is supervised learning

102
00:05:02,460 --> 00:05:04,330
还是无监督或者半监督
Or it is unsupervised, semi-supervised learning

103
00:05:04,380 --> 00:05:05,810
自监督学习的模型
or self-supervised models

104
00:05:06,860 --> 00:05:08,050
我们都统一到了
We all take them

105
00:05:08,620 --> 00:05:10,210
相同的符号体系下
Under the same notation

106
00:05:11,340 --> 00:05:12,210
我们会看到
You will see

107
00:05:12,580 --> 00:05:15,930
这些模型背后的历史关联和数学关联
the historical and mathematical relations among these models

108
00:05:18,430 --> 00:05:19,420
大家会有一个
I hope you will get

109
00:05:20,710 --> 00:05:23,860
总体的对自然语言处理建模的认识
an overall comprehension of NLP models

110
00:05:27,100 --> 00:05:28,570
那么第六个特点呢
The last feature is that

111
00:05:29,300 --> 00:05:32,050
就是课程的设置尽量允许灵活性
We want the course to be as flexible as possible

112
00:05:33,260 --> 00:05:35,570
不同的同学可以根据自己的需要
Students with different needs

113
00:05:36,180 --> 00:05:37,650
和自己所处的年级
and of different grades

114
00:05:38,140 --> 00:05:40,170
来选择所学习的内容
can get what they need accordingly

115
00:05:42,680 --> 00:05:45,030
那比如有的同学他只想了解自然
For example, if you only want to know

116
00:05:45,030 --> 00:05:46,850
语言处理都研究什么问题
what are the problems that NLP deals with

117
00:05:48,040 --> 00:05:50,450
事实上在第一大部分内容
you can find the answers in the first chapter

118
00:05:50,480 --> 00:05:52,610
我们将集中的讨论这些问题
where we will tackle on these problems intensively

119
00:05:56,600 --> 00:05:58,150
我们这课程呢
The textbook

120
00:06:00,000 --> 00:06:01,430
最主要的教材呢
for our course

121
00:06:03,480 --> 00:06:05,270
是由我和腾志杨博士
is Natural Language Processing - A Machine Learning perspective

122
00:06:06,080 --> 00:06:06,830
撰写的
Written by me

123
00:06:08,040 --> 00:06:09,150
自然语言处理
And

124
00:06:09,480 --> 00:06:10,110
这本书
Dr. Zhiyang Teng

125
00:06:12,250 --> 00:06:13,680
相关的一些资料呢
You can find it

126
00:06:14,530 --> 00:06:16,600
也可以在剑桥大学出版社的
from the Cambridge University Press

127
00:06:17,170 --> 00:06:18,600
网站上找到
website

128
00:06:23,680 --> 00:06:24,770
我们还推荐
We also recommend

129
00:06:26,760 --> 00:06:27,050
由
the book

130
00:06:28,800 --> 00:06:30,850
Dan Jurafsky 博士
Written by Dr. Dan Jurafsky

131
00:06:31,600 --> 00:06:33,810
和 James Martin 博士所
and Dr. James Martin

132
00:06:34,280 --> 00:06:35,650
共同撰写的这本
named

133
00:06:36,320 --> 00:06:38,450
Speech and Language Processing
Speech and Language Processing

134
00:06:40,390 --> 00:06:41,220
这本书呢
This book

135
00:06:42,470 --> 00:06:45,020
针对每个自然语言处理任务本身
gives elaborate discussions

136
00:06:46,310 --> 00:06:48,260
尤其是任务的语言学特点
to the nlp tasks itself

137
00:06:48,990 --> 00:06:51,180
和针对这个任务的重要特征
especially its linguistic characteristics

138
00:06:52,430 --> 00:06:54,260
进行了较为详细的讨论
And its significant features

139
00:06:56,370 --> 00:06:57,560
他的组织方式
The arrangement of their book

140
00:06:58,490 --> 00:07:00,720
和我们这门课程的主要教材
has certain difference

141
00:07:01,730 --> 00:07:02,680
有一定的不同
to our book

142
00:07:03,450 --> 00:07:05,280
可以作为补充来学习
Which can serve as an useful supplementary

143
00:07:09,160 --> 00:07:09,810
另外呢
And also

144
00:07:10,760 --> 00:07:11,530
我们推荐
We recommend

145
00:07:13,120 --> 00:07:15,170
宗成庆教授所书写的
the book written by Chengqing Zong

146
00:07:16,360 --> 00:07:18,430
统计自然语言处理这本书
Tong Ji Zi Ran Yu Yan Chu Li

147
00:07:19,240 --> 00:07:20,830
作为中文的
As a Chinese

148
00:07:21,640 --> 00:07:22,110
教材
Textbook

149
00:07:23,990 --> 00:07:27,240
大家可以看到各个自然语言处理任务
You will learn the NLP tasks

150
00:07:27,950 --> 00:07:29,200
的详细方法
in detail

151
00:07:29,750 --> 00:07:30,720
的中文介绍
and in Chinese

152
00:07:32,875 --> 00:07:35,225
这节课就上到这里咱们下次再见
This ends our course introduction, see you on the next