0
1
00:00:00,510 --> 00:00:04,400
In the next couple of lessons I want to show you two things.
1

2
00:00:04,440 --> 00:00:11,520
First I want to show you how to install third party modules and packages that don't come bundled with
2

3
00:00:11,520 --> 00:00:14,370
the Anaconda distribution that we're running.
3

4
00:00:14,610 --> 00:00:20,640
But more importantly, I want to show you a data visualization technique for giving prominence to words
4

5
00:00:20,730 --> 00:00:29,980
that appear more frequently in a piece of text, namely the word cloud. A word cloud or a tag cloud as
5

6
00:00:29,980 --> 00:00:33,420
it is also called looks like this.
6

7
00:00:33,520 --> 00:00:39,070
The idea behind a word cloud is that the words that are larger appear more frequently than those that
7

8
00:00:39,070 --> 00:00:40,740
are smaller. Here,
8

9
00:00:40,900 --> 00:00:46,240
I've used the Wikipedia article on machine learning to generate this particular word cloud.
9

10
00:00:46,240 --> 00:00:49,690
Notice how the larger tags catch your attention.
10

11
00:00:49,810 --> 00:00:56,530
You'll probably find that you're not actually reading all the words in this visualization at all.
11

12
00:00:56,530 --> 00:01:00,850
Instead people just tend to scan these word clouds.
12

13
00:01:00,850 --> 00:01:03,420
So, why use this technique?
13

14
00:01:03,460 --> 00:01:07,390
It's not scientific and it's not accurate.
14

15
00:01:07,540 --> 00:01:16,090
True, but more often than not, you need to grab people's attention and a word cloud can do this rather
15

16
00:01:16,090 --> 00:01:16,840
well.
16

17
00:01:17,080 --> 00:01:19,810
It's a little bit of a novel visualization.
17

18
00:01:19,810 --> 00:01:26,820
It's a bit creative and it communicates the relative importance of the words in the dataset very quickly.
18

19
00:01:27,220 --> 00:01:34,360
After all it's pretty easy to see that a word that's twice the size of another word is probably twice
19

20
00:01:34,360 --> 00:01:35,360
as important.
20

21
00:01:35,560 --> 00:01:39,610
And in our case it appears twice as often.
21

22
00:01:39,700 --> 00:01:45,670
Speaking of grabbing people's attention, you can get even more creative with the word cloud by applying
22

23
00:01:45,760 --> 00:01:49,110
a mask to the image. Here,
23

24
00:01:49,240 --> 00:01:54,360
I've used a circle as a mask, but you could use another image instead.
24

25
00:01:54,400 --> 00:02:01,540
For example, would you like to venture a guess as to what data I've used to create this word cloud in
25

26
00:02:01,540 --> 00:02:02,980
the shape of a whale?
26

27
00:02:03,340 --> 00:02:12,010
I actually used Herman Melville's novel, Moby Dick, and here I've used the data from our spam emails, and
27

28
00:02:12,010 --> 00:02:15,440
here I've used Shakespeare's play Hamlet.
28

29
00:02:15,460 --> 00:02:20,650
Coming up, I'll be showing you how to create all of these three visualizations.
29

30
00:02:20,650 --> 00:02:24,980
That way you can apply this technique to your future projects.
30

31
00:02:25,000 --> 00:02:31,360
I've seen a lot of people use word clouds on reports and presentations that say contain customer feedback
31

32
00:02:31,720 --> 00:02:38,620
or metadata like tags or SEO keywords that will help you rank on Google or really any situation where
32

33
00:02:38,620 --> 00:02:44,560
you quickly need to show how people feel about your product or your business.
33

34
00:02:44,560 --> 00:02:49,080
These are the kind of applications where word clouds really shine.
34

35
00:02:49,100 --> 00:02:49,500
All right.
35

36
00:02:49,510 --> 00:02:55,890
So how do we create a word cloud in Python? Because we're not gonna do it from scratch.
36

37
00:02:55,930 --> 00:03:00,230
We will use an existing Python package.
37

38
00:03:00,230 --> 00:03:06,790
Now the fantastic thing about the Anaconda distribution is that it comes bundled with lots and lots
38

39
00:03:06,790 --> 00:03:07,940
of different packages
39

40
00:03:07,960 --> 00:03:11,900
straight out of the box. These included numpy, pandas,
40

41
00:03:11,960 --> 00:03:15,310
matplotlib and many others that we've already used.
41

42
00:03:15,670 --> 00:03:22,450
However, the Python ecosystem is actually huge and you'll often find that you'll want to make use of
42

43
00:03:22,450 --> 00:03:26,390
a package that's not installed by default.
43

44
00:03:26,460 --> 00:03:34,140
In that case, you'll need to know how to download and install a custom package to add to your collection.
44

45
00:03:34,500 --> 00:03:41,310
For that, you'll be using the terminal if you're running a Mac or the Anaconda prompt on Windows.
45

46
00:03:41,310 --> 00:03:46,410
Both of these kind of look like this and this is where you can execute many different types of commands.
46

47
00:03:47,070 --> 00:03:52,760
The command that we'll be looking at in detail is called "conda install"
47

48
00:03:52,780 --> 00:04:00,480
and the easiest way to install a package is just to type "conda install" followed by the package name.
48

49
00:04:00,480 --> 00:04:08,810
However, this won't always work and it actually won't work in the case of our word cloud package.
49

50
00:04:08,970 --> 00:04:16,020
And the reason is is that we also have to specify something called a channel. A channel is the location
50

51
00:04:16,320 --> 00:04:19,630
where conda will look for the package.
51

52
00:04:20,040 --> 00:04:25,920
If you just write "conda install" and then the package name, then conda will search in a default location
52

53
00:04:26,190 --> 00:04:29,370
which may or may not contain the package that you're looking for.
53

54
00:04:29,730 --> 00:04:36,090
If you need to specify a particular channel or you know that the package exists, then all you have to
54

55
00:04:36,090 --> 00:04:45,260
do is follow "conda install" with "-c" and then the channel name. This syntax, the "-c" works just
55

56
00:04:45,260 --> 00:04:48,190
like you're supplying an argument to a Python function.
56

57
00:04:49,270 --> 00:04:57,610
In our case, the wordcloud package is located in a channel called "conda-forge" and therefore the code
57

58
00:04:57,820 --> 00:05:06,220
that we will be typing into our prompt will read "conda install -c conda-forge wordcloud".
58

59
00:05:06,250 --> 00:05:12,170
Let me quickly show you how to do this on Mac and then I'll show you what you'll see on a Windows machine.
59

60
00:05:12,960 --> 00:05:17,220
To see the existing packages that are already installed
60

61
00:05:17,220 --> 00:05:25,210
fire up your Anaconda navigator and then click on Environments. There you can search for the existing
61

62
00:05:25,210 --> 00:05:28,760
packages that are already installed on your computer.
62

63
00:05:28,900 --> 00:05:34,450
The package that we're looking for is called "wordcloud", all in one word.
63

64
00:05:34,450 --> 00:05:40,130
Now since I can't find it, I'm going to open my terminal. There
64

65
00:05:40,210 --> 00:05:48,460
I'm going to type "conda install wordcloud", but just as we've discussed earlier for the wordcloud package
65

66
00:05:48,640 --> 00:05:56,320
we actually need to specify a channel, we need to specify a location because it's not found in the default
66

67
00:05:56,320 --> 00:05:59,230
location where Conda is looking.
67

68
00:05:59,230 --> 00:06:07,750
Instead we need to type "conda install -c conda-forge wordcloud" and if you don't make
68

69
00:06:07,780 --> 00:06:13,990
any typos, then you should see your computer starting to do some work. Prior to going ahead with the installation
69

70
00:06:14,350 --> 00:06:21,190
you may have to confirm that you actually want to do this and then you have to type "y" for yes to proceed.
70

71
00:06:21,790 --> 00:06:24,540
At this point a download will start
71

72
00:06:24,790 --> 00:06:29,860
and once the download is finished, you'll see something like "Preparing transaction", "Verifying transaction",
72

73
00:06:30,250 --> 00:06:31,870
"Executing transaction".
73

74
00:06:31,870 --> 00:06:33,130
*Drum roll*
74

75
00:06:33,130 --> 00:06:37,230
And then you're good to go. Once you see these messages
75

76
00:06:37,230 --> 00:06:45,170
I encourage you to fire up Finder, go to the Anaconda folder, then go to the packages folder and then
76

77
00:06:45,170 --> 00:06:48,240
find the wordcloud subfolder there.
77

78
00:06:48,260 --> 00:06:53,720
This is the package that you've just installed. Back in the Anaconda Navigator,
78

79
00:06:53,810 --> 00:06:58,430
you can now type in "wordcloud" into the search bar and it should pop up.
79

80
00:06:58,490 --> 00:07:04,370
Mind you, I actually had to restart my Anaconda Navigator for it to pick this up. And that pretty much
80

81
00:07:04,370 --> 00:07:07,570
concludes the setup for Mac. On Windows,
81

82
00:07:07,580 --> 00:07:14,420
the story is pretty much the same. If you fire up the Anaconda Navigator and search for "wordcloud" it
82

83
00:07:14,420 --> 00:07:18,440
doesn't appear. To bring up the Anaconda prompt,
83

84
00:07:18,560 --> 00:07:24,240
all you need to do is click on that little place symbol and then say "Open Terminal".
84

85
00:07:24,340 --> 00:07:31,870
This will fire up the Anaconda prompt where you can type in "conda install -c conda-forge
85

86
00:07:32,950 --> 00:07:34,340
wordcloud".
86

87
00:07:34,530 --> 00:07:40,900
Now depending on the speed of your machine, this might take a little while, but you can't go away and grab
87

88
00:07:40,900 --> 00:07:48,280
a coffee just yet, because once again you need to type in "y" to confirm and proceed. After you've done
88

89
00:07:48,280 --> 00:07:50,230
that a download will start
89

90
00:07:50,460 --> 00:07:56,470
and I'm actually fast forwarding my video here because this took a little bit longer on this machine,
90

91
00:07:57,040 --> 00:08:02,740
but at the end you'll see the very same messages that we saw on Mac - "Preparing transaction", "Verifying
91

92
00:08:02,740 --> 00:08:05,290
transaction", "Executing transaction".
92

93
00:08:05,290 --> 00:08:06,060
*Drum roll*
93

94
00:08:06,700 --> 00:08:07,810
And that's it.
94

95
00:08:08,620 --> 00:08:15,160
If you'd like to see where these packages were actually installed on your computer, go to your user directory,
95

96
00:08:15,160 --> 00:08:17,240
so mine's called "P.D"
96

97
00:08:17,470 --> 00:08:20,880
under Users, but yours will have your own name of course.
97

98
00:08:21,040 --> 00:08:22,510
Then open Anaconda
98

99
00:08:22,510 --> 00:08:27,490
3, then go to the packages folder and there,
99

100
00:08:27,600 --> 00:08:32,850
if you scroll further down you should see the wordcloud package along with the others that start with
100

101
00:08:32,850 --> 00:08:36,000
"w". Back in the Anaconda Navigator,
101

102
00:08:36,000 --> 00:08:40,260
you won't necessarily find the wordcloud package listed already,
102

103
00:08:40,260 --> 00:08:46,290
you might have to actually quit and restart this thing, because only then does it seem to pick it up.
103

104
00:08:46,560 --> 00:08:51,390
Now you're pretty much all set to install any package you come across in the future.
104

105
00:08:51,690 --> 00:08:59,430
All you need to do after all is type "conda install", then the channel where the package is located and
105

106
00:08:59,430 --> 00:09:04,380
then the package name and that's it. In the next lesson,
106

107
00:09:04,380 --> 00:09:08,550
it's time to write some Python code and actually create these word clouds.
107

108
00:09:08,550 --> 00:09:11,180
So I hope you're looking forward to that as much as I am.
108

109
00:09:11,190 --> 00:09:12,320
I'll see you there.