1
00:00:00,710 --> 00:00:01,620
- [Instructor] So to finish up

2
00:00:01,620 --> 00:00:04,260
our trending topics discussion

3
00:00:04,260 --> 00:00:09,060
let's take the top trending
topics from New York City

4
00:00:09,060 --> 00:00:13,260
and turn them into a
word cloud visualization.

5
00:00:13,260 --> 00:00:15,480
So for that purpose we're going to define

6
00:00:15,480 --> 00:00:19,630
a dictionary called topics
and we're going to be using

7
00:00:19,630 --> 00:00:22,730
this dictionary to store key value pairs

8
00:00:22,730 --> 00:00:25,980
where the keys are the
trending topic texts

9
00:00:25,980 --> 00:00:29,130
and the values are the number of tweets

10
00:00:29,130 --> 00:00:31,810
associated with that trending topic.

11
00:00:31,810 --> 00:00:33,810
And remember in a word cloud

12
00:00:33,810 --> 00:00:37,190
the visualization is
going to make bigger words

13
00:00:37,190 --> 00:00:40,210
for the items that are more frequent.

14
00:00:40,210 --> 00:00:44,510
So to create the dictionary
let's use a little loop here.

15
00:00:44,510 --> 00:00:45,640
We're going to walk through

16
00:00:45,640 --> 00:00:48,050
the New York City trends
list that we created

17
00:00:48,050 --> 00:00:51,130
back up above here in the preceding video.

18
00:00:51,130 --> 00:00:55,000
And for each of the items in that list

19
00:00:55,000 --> 00:00:57,840
we're going to create a new key

20
00:00:57,840 --> 00:01:01,400
in the topics dictionary,
which is initially empty.

21
00:01:01,400 --> 00:01:05,820
The key is going to be
the given trends name

22
00:01:05,820 --> 00:01:10,820
so that will often be a hashtag
or just a string of text.

23
00:01:10,920 --> 00:01:13,010
And the value is going to be

24
00:01:13,010 --> 00:01:15,940
the corresponding trends tweet volume.

25
00:01:15,940 --> 00:01:18,130
And remember that up above here

26
00:01:18,130 --> 00:01:20,810
we filtered out any of the items

27
00:01:20,810 --> 00:01:23,990
that have none as their tweet volume.

28
00:01:23,990 --> 00:01:25,950
So all of the tweets that are going to be

29
00:01:25,950 --> 00:01:28,790
processed in this case are going to have

30
00:01:28,790 --> 00:01:32,080
10,000 or more tweets
associated with them.

31
00:01:32,080 --> 00:01:34,560
All of the trends will
have 10,000 or more tweets

32
00:01:34,560 --> 00:01:38,320
and because it's a relatively small number

33
00:01:38,320 --> 00:01:40,700
of topics that we're dealing with

34
00:01:40,700 --> 00:01:43,000
the word cloud is going to probably be

35
00:01:43,000 --> 00:01:46,360
somewhat sparse when it is completed.

36
00:01:46,360 --> 00:01:48,340
So let's go ahead and execute that

37
00:01:48,340 --> 00:01:50,320
to acquire the key value pairs.

38
00:01:50,320 --> 00:01:53,890
And we can take a quick look
at what topics looks like

39
00:01:53,890 --> 00:01:56,540
and you'll see that we actually got

40
00:01:56,540 --> 00:01:59,050
a decent number of key value pairs.

41
00:01:59,050 --> 00:02:03,690
The top tweet topic at the
moment is whatever Hanbin is

42
00:02:03,690 --> 00:02:07,780
and there's over 2.4 million
tweets for that right now

43
00:02:07,780 --> 00:02:11,010
and that goes all the
way down to the smallest

44
00:02:11,010 --> 00:02:13,430
of the topics which has
as of right this moment

45
00:02:13,430 --> 00:02:16,280
11,000 plus tweets today.

46
00:02:16,280 --> 00:02:19,350
And there are some other
trending topics probably as well

47
00:02:19,350 --> 00:02:23,000
but those were fewer than 10,000 tweets

48
00:02:23,000 --> 00:02:25,470
and therefore did not show up in the list.

49
00:02:25,470 --> 00:02:28,520
And always keep in mind
that when you look at stuff

50
00:02:28,520 --> 00:02:30,980
like this coming back live from Twitter

51
00:02:30,980 --> 00:02:34,900
you could get some really crazy stuff back

52
00:02:34,900 --> 00:02:37,380
including lots of curse
words, for example.

53
00:02:37,380 --> 00:02:41,350
So just be aware that you
may see things you don't like

54
00:02:41,350 --> 00:02:43,650
in the results that you
get back from Twitter.

55
00:02:44,520 --> 00:02:48,290
So at this point we have our dictionary

56
00:02:48,290 --> 00:02:50,800
that we're going to use
to create the word cloud.

57
00:02:50,800 --> 00:02:52,900
Previously we didn't use a dictionary

58
00:02:52,900 --> 00:02:54,700
so this is a little bit different

59
00:02:54,700 --> 00:02:57,950
from what we showed you
back in the NLP chapter.

60
00:02:57,950 --> 00:03:01,390
So next up we're going to
import the word cloud class

61
00:03:01,390 --> 00:03:02,960
from the word cloud library

62
00:03:02,960 --> 00:03:05,530
and this time when I create the word cloud

63
00:03:05,530 --> 00:03:09,340
I'm going to include some
additional key word arguments.

64
00:03:09,340 --> 00:03:11,010
We do have a width and a height

65
00:03:11,010 --> 00:03:12,730
so we're specifying that we want

66
00:03:12,730 --> 00:03:15,920
600 pixels wide and 900 pixels tall.

67
00:03:15,920 --> 00:03:18,270
We're not going to use
a mask image this time

68
00:03:18,270 --> 00:03:23,140
so our actual word cloud
will be rectangular

69
00:03:23,140 --> 00:03:26,070
and using the dimensions
that we've specified.

70
00:03:26,070 --> 00:03:29,390
This key word argument is
new to us in this example

71
00:03:29,390 --> 00:03:32,770
prefer horizontal with the value .5

72
00:03:32,770 --> 00:03:36,330
means to try to use as a guideline

73
00:03:36,330 --> 00:03:41,270
that 50% of the words should
be horizontally oriented

74
00:03:41,270 --> 00:03:43,640
versus vertically oriented.

75
00:03:43,640 --> 00:03:46,189
And it's free to ignore that

76
00:03:46,189 --> 00:03:49,200
in order to fit the content appropriately

77
00:03:49,200 --> 00:03:52,740
but it's just a suggestion
that you can provide.

78
00:03:52,740 --> 00:03:54,420
We also indicated that we didn't want

79
00:03:54,420 --> 00:03:58,120
any fonts smaller than 10 point size.

80
00:03:58,120 --> 00:04:01,140
And we did give a color map
to choose colors in this case

81
00:04:01,140 --> 00:04:04,900
and we set the background
color to white as well.

82
00:04:04,900 --> 00:04:07,910
Now once we've created
the word cloud object

83
00:04:07,910 --> 00:04:10,770
we can use it to generate the word cloud.

84
00:04:10,770 --> 00:04:13,610
Previously we used the
generate method for that,

85
00:04:13,610 --> 00:04:15,660
here we're using fit words

86
00:04:15,660 --> 00:04:18,060
which is capable of
receiving as its argument

87
00:04:18,060 --> 00:04:20,310
a dictionary of key value pairs

88
00:04:20,310 --> 00:04:22,890
like the one that we created up above.

89
00:04:22,890 --> 00:04:24,950
So we'll go ahead and execute that

90
00:04:24,950 --> 00:04:28,630
to create the word cloud and
let's also write that out

91
00:04:29,610 --> 00:04:32,460
to a file on disk as a PNG image

92
00:04:32,460 --> 00:04:36,190
by using the to file method
of the word cloud object.

93
00:04:36,190 --> 00:04:38,990
And I'm now going to go
ahead and open up that image

94
00:04:38,990 --> 00:04:40,910
and bring it onto my screen here.

95
00:04:40,910 --> 00:04:43,560
So this is, whoops, this is the word cloud

96
00:04:43,560 --> 00:04:47,310
that we just generated and as you can see

97
00:04:47,310 --> 00:04:50,650
some of the words are horizontal
some of them are vertical.

98
00:04:50,650 --> 00:04:53,040
The one that's the biggest is the one

99
00:04:53,040 --> 00:04:55,070
that had the largest tweet count

100
00:04:55,070 --> 00:04:57,440
and these smaller items that you see

101
00:04:57,440 --> 00:05:00,510
in various locations
throughout the word cloud

102
00:05:00,510 --> 00:05:04,683
represent the ones that had
smaller overall tweet counts.