﻿1
00:00:00,166 --> 00:00:02,533
Hello and welcome to this art tutorial.

2
00:00:02,533 --> 00:00:05,466
So we're still trying as hard
as we can to simplify

3
00:00:05,466 --> 00:00:09,166
the corpus in order to reduce the future
sparse matrix of features.

4
00:00:09,166 --> 00:00:10,666
As much as possible.

5
00:00:10,666 --> 00:00:13,466
And so far
we put all the words in lowercase.

6
00:00:13,466 --> 00:00:16,700
We removed all the numbers
and removed all the punctuation.

7
00:00:16,933 --> 00:00:18,333
And in today's tutorial

8
00:00:18,333 --> 00:00:21,866
we are going to remove
all the non relevant words of our reviews.

9
00:00:22,200 --> 00:00:23,766
So what is the non relevant word.

10
00:00:23,766 --> 00:00:26,466
It's for example this word here.

11
00:00:26,466 --> 00:00:31,300
Indeed if we know that this first review
well love this place is a positive review.

12
00:00:31,400 --> 00:00:32,833
It's not thanks to this word.

13
00:00:32,833 --> 00:00:35,833
It's of course thanks to the loved word.

14
00:00:35,866 --> 00:00:39,633
So that means that this word here
is not a relevant word.

15
00:00:39,833 --> 00:00:43,633
And so it's totally non relevant for us
and not useful to include it

16
00:00:43,633 --> 00:00:45,433
in the future. Sparse matrix.

17
00:00:45,433 --> 00:00:48,166
That will be nothing else
than the matrix of features

18
00:00:48,166 --> 00:00:50,900
and the outcome whether the review
is positive or negative.

19
00:00:50,900 --> 00:00:52,800
So that's what we'll do in this tutorial.

20
00:00:52,800 --> 00:00:55,566
We will remove
all these non relevant words.

21
00:00:55,566 --> 00:00:59,166
And we will update our corpus of reviews
by removing all these words.

22
00:00:59,600 --> 00:01:01,333
All right. So same as before.

23
00:01:01,333 --> 00:01:06,133
Very simply
we take this line copy and paste it below.

24
00:01:06,400 --> 00:01:09,200
And in this line we're going to replace
remove

25
00:01:09,200 --> 00:01:12,600
punctuation by remove words.

26
00:01:13,933 --> 00:01:15,100
But that's not all this time.

27
00:01:15,100 --> 00:01:16,900
It's not as simple as before.

28
00:01:16,900 --> 00:01:18,600
We need to add something

29
00:01:18,600 --> 00:01:22,200
which is a something that will specify
which words we want to remove.

30
00:01:22,500 --> 00:01:25,866
And actually there is a built
in list of non relevant words

31
00:01:26,100 --> 00:01:28,066
that is called stopwords.

32
00:01:28,066 --> 00:01:31,500
And that basically contains
all the non relevant words like this.

33
00:01:31,800 --> 00:01:36,500
Although the articles prepositions
like and or well you know all these words

34
00:01:36,500 --> 00:01:38,033
that don't help the machine

35
00:01:38,033 --> 00:01:41,066
learning algorithm figure out
if they reduce positive or negative.

36
00:01:41,533 --> 00:01:46,366
So this list of non relevant
words in the Stopwords list is practically

37
00:01:46,366 --> 00:01:50,033
always used in natural language
processing, because indeed, these words

38
00:01:50,033 --> 00:01:54,066
will never help you or your algorithm
to classify your texts.

39
00:01:54,300 --> 00:01:56,400
So you will most of the time use it.

40
00:01:56,400 --> 00:01:59,600
And therefore that's a very important step
because of course

41
00:01:59,600 --> 00:02:03,733
that simplifies the corpus and reduces
the future sparse matrix very much.

42
00:02:03,933 --> 00:02:08,400
So as I just said, this is not what we
only need to input in this map function.

43
00:02:08,666 --> 00:02:10,866
We need to input a third parameter.

44
00:02:10,866 --> 00:02:13,733
And that corresponds to the words
we want to remove.

45
00:02:13,733 --> 00:02:15,633
That is all the non relevant words.

46
00:02:15,633 --> 00:02:18,866
And these words are in this stopwords

47
00:02:20,000 --> 00:02:21,233
function.

48
00:02:21,233 --> 00:02:22,200
All right.

49
00:02:22,200 --> 00:02:26,600
So basically this returns all the words
that are not relevant for our model.

50
00:02:26,733 --> 00:02:29,700
And therefore thanks to this function here
we will remove

51
00:02:29,700 --> 00:02:32,700
all the words returned by this Stopwords
function.

52
00:02:32,866 --> 00:02:33,233
All right.

53
00:02:33,233 --> 00:02:35,000
So that's all for this line of code.

54
00:02:35,000 --> 00:02:38,300
But we need to add a little something.

55
00:02:38,333 --> 00:02:39,866
Well especially for you

56
00:02:39,866 --> 00:02:42,866
if you're doing natural language
processing for the first time,

57
00:02:43,033 --> 00:02:46,833
which is a library that we need to install
and import

58
00:02:47,000 --> 00:02:51,466
to be able to use the stop words function,
because this function is not included

59
00:02:51,466 --> 00:02:53,100
in the default package of R,

60
00:02:53,100 --> 00:02:56,100
so we need to install the required package
to use this function.

61
00:02:56,400 --> 00:02:58,533
And this package
actually has a funny name.

62
00:02:58,533 --> 00:03:01,200
It is called snowball C.

63
00:03:01,200 --> 00:03:06,733
And so let's now install this package
so we can, you know, copy this line,

64
00:03:07,500 --> 00:03:08,333
paste it here.

65
00:03:08,333 --> 00:03:13,900
And in this quote here in the parenthesis
we input snowball.

66
00:03:13,900 --> 00:03:16,900
So it's spelled this way. Snowball.

67
00:03:16,966 --> 00:03:18,533
And then C.

68
00:03:18,533 --> 00:03:18,833
All right.

69
00:03:18,833 --> 00:03:22,800
So right now in some comments but
you know check in your package this list

70
00:03:22,800 --> 00:03:25,933
if you already have this snowball package
we never know.

71
00:03:26,400 --> 00:03:29,500
And if you don't have it
well you can execute this line

72
00:03:29,500 --> 00:03:32,500
without the command
to install the package.

73
00:03:32,533 --> 00:03:33,366
All right.

74
00:03:33,366 --> 00:03:37,166
And now of course, as usual
we will import the package

75
00:03:37,200 --> 00:03:40,000
automatically
thanks to this library function.

76
00:03:40,000 --> 00:03:44,033
So we will same copy this line, paste

77
00:03:44,066 --> 00:03:49,533
it below and replace
time by actually snowball

78
00:03:50,733 --> 00:03:52,766
C. All right snowball.

79
00:03:52,766 --> 00:03:56,000
See now the required package
are installed

80
00:03:56,000 --> 00:03:59,133
and imported to be able to use this
Stopwords function.

81
00:03:59,366 --> 00:04:01,200
So everything is all good.

82
00:04:01,200 --> 00:04:04,266
And now let's try it on the first review.

83
00:04:04,266 --> 00:04:08,400
Because the first review contains
some irrelevant words like this.

84
00:04:08,400 --> 00:04:12,433
Here, it might be the only word
that is removed because you know

85
00:04:12,433 --> 00:04:16,700
well might not be a word of this
stopwords list because, you know,

86
00:04:16,700 --> 00:04:21,500
the Stopwords list is a list of common
words like, articles and prepositions,

87
00:04:21,766 --> 00:04:25,300
common words, but irrelevant words
and well, is not common.

88
00:04:25,300 --> 00:04:26,966
So it might not be removed.

89
00:04:26,966 --> 00:04:30,100
But definitely this will be removed
because this is a common

90
00:04:30,300 --> 00:04:31,833
and non relevant word.

91
00:04:31,833 --> 00:04:33,133
So let's check it out.

92
00:04:33,133 --> 00:04:36,433
We will select this line and execute.

93
00:04:36,833 --> 00:04:37,533
Here we go.

94
00:04:37,533 --> 00:04:41,700
And now let's have a look at
the first review by pressing the up arrow.

95
00:04:42,066 --> 00:04:44,266
Here it is. Let's press enter.

96
00:04:44,266 --> 00:04:47,266
And as I just said this was removed.

97
00:04:47,700 --> 00:04:51,166
So now the first review becomes
well loved place.

98
00:04:51,533 --> 00:04:54,900
And you know
even if we really simplify the review no.

99
00:04:54,900 --> 00:04:57,333
Now it doesn't
look the same as the original review.

100
00:04:57,333 --> 00:05:00,700
Well we can still understand
that it's a positive review.

101
00:05:00,700 --> 00:05:04,333
And especially our machine learning
model will understand it very well.

102
00:05:04,600 --> 00:05:08,033
And that will be
thanks to this loved word here,

103
00:05:08,133 --> 00:05:11,833
which is a word that, of course,
might be present in some other reviews,

104
00:05:12,066 --> 00:05:15,066
which themselves will be positive
reviews as well.

105
00:05:15,233 --> 00:05:18,600
So that's how our machine learning
algorithm will understand

106
00:05:18,733 --> 00:05:22,900
that love indicates a positive review,
and therefore

107
00:05:22,900 --> 00:05:26,900
that's all it needs
to establish that kind of correlations.

108
00:05:26,900 --> 00:05:29,633
And this is totally unused form.

109
00:05:29,633 --> 00:05:31,900
And we were right to remove it.

110
00:05:31,900 --> 00:05:32,266
All right.

111
00:05:32,266 --> 00:05:33,900
So that's done for this step.

112
00:05:33,900 --> 00:05:35,800
That was another very important step.

113
00:05:35,800 --> 00:05:36,733
But that's not all.

114
00:05:36,733 --> 00:05:37,666
In the next tutorial

115
00:05:37,666 --> 00:05:41,800
we'll do another very important step
which is the stemming step.

116
00:05:42,200 --> 00:05:43,666
And so I'll explain you what it is

117
00:05:43,666 --> 00:05:47,266
and how we perform this new cleaning
step in the next tutorial.

118
00:05:47,566 --> 00:05:49,166
Until then, enjoy machine learning.
