0
1
00:00:00,500 --> 00:00:01,320
Are you ready?
1

2
00:00:01,320 --> 00:00:03,130
Here's the solution.
2

3
00:00:03,390 --> 00:00:13,030
Going to the top of the file, I'll add another constant and I'll call this constant "SKULL_FILE" and it's going
3

4
00:00:13,030 --> 00:00:18,180
to point to the "skull-icon.png",
4

5
00:00:21,190 --> 00:00:31,290
this one right here, Shift+Enter on this cell, scroll back down and now it's time to put all of this text
5

6
00:00:31,530 --> 00:00:39,180
that's in the "shakespeare-hamlet.txt" file from the NLTK resources into a single string,
6

7
00:00:40,710 --> 00:00:53,610
so I'll call a variable "hamlet_corpus" and set that equal to a "nltk.corpus.gutenberg.
7

8
00:00:53,850 --> 00:01:04,890
words('shakespeare-hamlet.txt')" the same file name as right
8

9
00:01:04,890 --> 00:01:17,740
here. The word list is going to be equal to "[''.join()]" and I'll join the
9

10
00:01:17,740 --> 00:01:31,030
word up from my loop, so "for word in hamlet_corpus" will give us our list of words and then I'll say "hamlet_
10

11
00:01:31,450 --> 00:01:44,570
as_string = ' '.join(word_list)". That's our play as a string.
11

12
00:01:44,920 --> 00:01:54,810
Now all I need to do is maybe go "skull_icon = Image.open()",
12

13
00:01:54,880 --> 00:02:02,550
so here we're using pillow, feeding the relative path to the file and now it's time to create that mask.
13

14
00:02:02,650 --> 00:02:15,730
So I'll say "image_mask" is equal to a new pillow image object with mode equal to "RGB",
14

15
00:02:16,810 --> 00:02:33,640
the size equal to the "skull_icon.size" and color equal to white - (255, 255, 255)" inside
15

16
00:02:33,700 --> 00:02:44,830
a tuple. The rgb_array for this skull image will simply be a numpy array created from the image mask.
16

17
00:02:47,190 --> 00:02:56,310
Now we can create our word cloud, "word_cloud" is gonna be equal to "WordCloud(
17

18
00:02:56,790 --> 00:03:10,310
mask = rgb_array)" from Hamlet. I'm going to choose my background color as "white" and then I'm
18

19
00:03:10,310 --> 00:03:18,560
going to pick a color map. The color map that I'm going to go for for my skull image is gonna be called
19

20
00:03:18,860 --> 00:03:30,080
"bone". I figure that's gonna be a nice combination. And for the maximum words, "max_words" argument,
20

21
00:03:30,300 --> 00:03:40,080
I'm going to pick... I'll start out with 50 and then I'm going to up it, just so you can see how a higher
21

22
00:03:40,080 --> 00:03:51,090
number will actually look a lot better. Then I'll use "word_cloud.generate()" to generate my image, but
22

23
00:03:51,480 --> 00:03:54,920
of course I have to feed in which string I want to use,
23

24
00:03:54,930 --> 00:03:58,290
so "hamlet_as_string".
24

25
00:03:58,290 --> 00:03:59,910
That's it for the word cloud.
25

26
00:03:59,910 --> 00:04:01,050
Now it's all gonna be 
26

27
00:04:01,050 --> 00:04:07,230
matplotlib. "plt.figure()" sets the size of our figure,
27

28
00:04:07,230 --> 00:04:10,460
I'm gonna go again with 16 and 8.
28

29
00:04:10,470 --> 00:04:17,610
I think that's a good science for me and for the video. I'll use the "imshow()" method, feed in the word_
29

30
00:04:17,610 --> 00:04:26,130
cloud along with an interpolation of "bilinear",
30

31
00:04:26,510 --> 00:04:29,700
so I think that's spelled right.
31

32
00:04:29,910 --> 00:04:30,830
Good.
32

33
00:04:30,960 --> 00:04:35,310
I'll remove the axes, so "axis('off')"
33

34
00:04:39,040 --> 00:04:41,650
and then I'll show my chart.
34

35
00:04:41,710 --> 00:04:44,930
The moment of truth of course is hitting Shift+Enter on this.
35

36
00:04:45,010 --> 00:04:45,970
Let's see what it looks like.
36

37
00:04:51,600 --> 00:04:52,010
All right.
37

38
00:04:52,020 --> 00:04:57,740
So I've got a value error: "canvas size is too small".
38

39
00:04:57,870 --> 00:05:00,180
Any idea what went wrong?
39

40
00:05:00,180 --> 00:05:03,540
The reason is I've got an entirely white canvas.
40

41
00:05:03,660 --> 00:05:06,740
The RGB array is fully white.
41

42
00:05:06,750 --> 00:05:14,440
There is no place for the word cloud to draw anything and that's because I haven't called
42

43
00:05:14,580 --> 00:05:20,460
"image_mask.paste()" with the skull icon.
43

44
00:05:20,460 --> 00:05:22,780
So I'll need to write "image_mask.
44

45
00:05:22,840 --> 00:05:33,550
paste(skull_icon, box = skull_icon)" and if I refresh my cell now, then this error
45

46
00:05:33,640 --> 00:05:34,390
should disappear.
46

47
00:05:35,570 --> 00:05:46,090
So here's my image and nobody in the world will be able to tell that this is indeed a skull image.
47

48
00:05:46,160 --> 00:05:53,050
The culprit this time is the maximum number of words. This number here is too small.
48

49
00:05:53,060 --> 00:05:57,690
It doesn't actually help us see any of the detail in the image.
49

50
00:05:57,920 --> 00:06:03,070
So I'm gonna change it from 50 to 600 and refresh my cell.
50

51
00:06:04,320 --> 00:06:10,830
For this many words your computer might actually run for a little while because it's doing a lot of
51

52
00:06:10,830 --> 00:06:15,690
work to generate this image, but I think the wait was worth it.
52

53
00:06:15,690 --> 00:06:22,410
We've got a beautiful word cloud here with a fantastic color scheme and a lot of the details in the
53

54
00:06:22,410 --> 00:06:27,600
word cloud being made visible by the high number of words in the word cloud.
54

55
00:06:27,600 --> 00:06:34,770
Now that we're done with you to practice word clouds and using the NLTK resources it's time to create
55

56
00:06:34,890 --> 00:06:35,820
a word cloud
56

57
00:06:35,820 --> 00:06:44,390
for our ham and our spam messages. Those we will create from our dataset and I'll see you in the next lesson.
57

58
00:06:44,400 --> 00:06:44,940
Take care.