0
1
00:00:00,640 --> 00:00:00,990
All right.
1

2
00:00:01,020 --> 00:00:08,760
So in the last lesson, we looked at how we could use any search term that we can come up with to search
2

3
00:00:08,760 --> 00:00:17,580
Twitter using the Twitter API and Swifter to get a 100 tweets back that match or contain this particular
3

4
00:00:17,580 --> 00:00:18,750
search term.
4

5
00:00:18,930 --> 00:00:26,120
And then, we ran it through our sentimentClassifier to get a sentiment for each of the 100 tweets.
5

6
00:00:26,430 --> 00:00:34,080
And then, we created a score depending on how many positive or how many negative sentiments there were
6

7
00:00:34,140 --> 00:00:36,480
in that set of 100 tweets.
7

8
00:00:36,480 --> 00:00:42,210
Now, of course, the user can't go into our code and change a string manually.
8

9
00:00:42,210 --> 00:00:49,680
So in this lesson, we're going to make our app interface work and integrate our code so that the user
9

10
00:00:49,680 --> 00:00:56,730
will be able to type in a particular hashtag or a particular handle and press the predict button to
10

11
00:00:56,730 --> 00:01:02,840
generate an emoticon that represents how people feel about that particular topic
11

12
00:01:02,850 --> 00:01:04,370
over 100 tweets.
12

13
00:01:04,590 --> 00:01:06,610
So let's get started doing that.
13

14
00:01:06,690 --> 00:01:12,590
And the first thing we have to do is we need to tap into the text that's inside the textField.
14

15
00:01:12,840 --> 00:01:21,690
So that is, of course, simply denoted by something like, let's say, let's searchText = textField
15

16
00:01:21,780 --> 00:01:22,480
.text.
16

17
00:01:22,860 --> 00:01:28,880
And, remember, this is a optional string because the user might have left it blank.
17

18
00:01:28,920 --> 00:01:33,260
So let's go ahead and add this as a "if let"
18

19
00:01:33,600 --> 00:01:39,600
so that we can unwrap the optional only when the textField is not empty because we don't want to be
19

20
00:01:39,600 --> 00:01:43,910
making API requests with an empty searchText.
20

21
00:01:43,950 --> 00:01:50,740
And now, let's go ahead and include all of this code up to, but not including the last brackets,
21

22
00:01:50,940 --> 00:01:55,370
and let's put it inside our brand-new "if let."
22

23
00:01:55,620 --> 00:02:03,570
And now instead of using a hardcoded string, we can change it to our searchText.
23

24
00:02:03,600 --> 00:02:03,870
All right.
24

25
00:02:03,900 --> 00:02:10,890
So the next thing that we need to do is we need to be able to activate this search for tweets pinging
25

26
00:02:11,120 --> 00:02:19,890
the Twitter API as well as making the predictions all when the user presses the predict button.
26

27
00:02:19,890 --> 00:02:26,310
So we need to take all of this code that's currently inside viewDidLoad and we need to move it out
27

28
00:02:26,310 --> 00:02:30,730
of viewDidLoad and into our IBAction here.
28

29
00:02:30,930 --> 00:02:34,210
So we don't actually need anything to happen on
29

30
00:02:34,230 --> 00:02:35,480
viewDidLoad anymore.
30

31
00:02:35,490 --> 00:02:41,550
We only had it there in order to test our code so that every time we run it, it would carry it out and
31

32
00:02:41,550 --> 00:02:45,210
it would print the results into our debug console.
32

33
00:02:45,240 --> 00:02:50,890
So, now all of this gets carried out when we press the predict button instead.
33

34
00:02:50,940 --> 00:02:57,090
So when we press the predict button, we want to create a sentimentScore and we don't want to see it
34

35
00:02:57,180 --> 00:03:01,830
in the debug console because, remember, the user can't see any of that.
35

36
00:03:01,830 --> 00:03:10,110
Instead, we want to represent that sentimentScore as an emoji so that when the user hits predict
36

37
00:03:10,350 --> 00:03:17,280
and gets the results back, it should update this emoji to represent how positive or negative the sentiment
37

38
00:03:17,280 --> 00:03:18,570
was.
38

39
00:03:18,570 --> 00:03:22,800
So in order to do that, we can just go ahead and add an "if" statement.
39

40
00:03:22,860 --> 00:03:31,160
So let's say if sentimentScore is greater than 20, so that's pretty much overwhelmingly positive,
40

41
00:03:31,320 --> 00:03:41,130
then the sentimentLabel, it's text property, which, remember, has to be a string, and emojis are actually
41

42
00:03:41,130 --> 00:03:42,910
unicode characters.
42

43
00:03:43,020 --> 00:03:49,290
So all we need to do here is to pull up our emoticon chooser which you can do by using the shortcut
43

44
00:03:49,320 --> 00:03:56,910
controller command and space, or you can simply go into edit and emoji and symbols, it does exactly the same
44

45
00:03:56,910 --> 00:03:57,720
thing.
45

46
00:03:57,840 --> 00:04:05,490
And inside here, if it's really positive, I'm going to choose to show this emoji and, again, remember that
46

47
00:04:05,490 --> 00:04:12,130
the sentimentLabel is something that is at class level and we're currently inside a closure,
47

48
00:04:12,300 --> 00:04:19,710
so we have to mark this property with a "self." And then, I'm just gonna go through this and add a whole
48

49
00:04:19,710 --> 00:04:28,650
bunch of cases where if instead of the sentimentScore being greater than 20, what if it was simply greater
49

50
00:04:28,650 --> 00:04:29,470
than 10,
50

51
00:04:29,550 --> 00:04:34,320
so between 10 and 20. Because, remember, at this point, this check would have already been done,
51

52
00:04:34,320 --> 00:04:36,390
so it's definitely not greater than 20.
52

53
00:04:36,780 --> 00:04:44,070
But if it's greater than 10, then in that case, maybe I'll want the sentimentLabel to have an emoji
53

54
00:04:44,100 --> 00:04:50,100
that looks kind of happy but, obviously, less happy than the previous one.
54

55
00:04:50,100 --> 00:04:58,060
And I'm just going to go through this and add an emoji for most of the ranges.
55

56
00:04:58,260 --> 00:05:04,920
So, now I have a whole bunch of conditions for, say, if the sentimentScore was above 20, or if it's between
56

57
00:05:04,980 --> 00:05:15,060
10 and 20, between 0 and 10, or at--or equal to zero, or between minus 10 and zero, between minus 20 and
57

58
00:05:15,060 --> 00:05:16,320
minus 10.
58

59
00:05:16,320 --> 00:05:20,280
And finally, if it's even less than minus 20.
59

60
00:05:20,340 --> 00:05:22,590
So, now let's give our app a try.
60

61
00:05:25,620 --> 00:05:31,710
So let's try something like @CocaCola.
61

62
00:05:31,720 --> 00:05:34,470
Okay, so people are kind of happy about it.
62

63
00:05:34,470 --> 00:05:39,280
So this is gonna be something between zero and 10.
63

64
00:05:39,570 --> 00:05:41,790
Now, let's try their competitor.
64

65
00:05:41,790 --> 00:05:47,570
How do people feel about Pepsi? Interesting.
65

66
00:05:47,590 --> 00:05:52,710
So people are not so happy with Pepsi, but they seem to like Coca-Cola more.
66

67
00:05:52,720 --> 00:05:54,940
That's interesting.
67

68
00:05:54,940 --> 00:06:03,190
Now, what if we try some stock symbols? Let's say, AAPL for Apple. Ooh, not so happy.
68

69
00:06:03,190 --> 00:06:13,260
What about something like Twitter? TWTR. Okay, people are feeling kind of mad about it.
69

70
00:06:13,270 --> 00:06:17,060
What about Facebook?
70

71
00:06:17,120 --> 00:06:22,570
What about Microsoft?
71

72
00:06:22,620 --> 00:06:23,610
Interesting.
72

73
00:06:23,610 --> 00:06:26,490
So you could play with this all day long.
73

74
00:06:26,490 --> 00:06:29,430
And believe me, I've been playing with it all day long.
74

75
00:06:29,640 --> 00:06:38,250
But now, essentially, we have a very simple interface and our app is able to do a lot of heavy lifting
75

76
00:06:38,490 --> 00:06:45,870
all in the background using the powers of CoreML 2 and the Natural Language processing that it brings
76

77
00:06:45,870 --> 00:06:50,590
us, as well as the machine learning model that we build ourselves
77

78
00:06:50,610 --> 00:06:58,960
of only a thousand tweets, and their sentiment analysis. And we've managed to use the Twitter API to pull
78

79
00:06:58,960 --> 00:07:07,470
live tweet data and filter it by a variety of means in order to generate this hundred tweet
79

80
00:07:07,800 --> 00:07:12,390
sentimentScore which we're using to power our little app.
80

81
00:07:12,390 --> 00:07:18,990
So I hope you can see that the possibilities for this app are absolutely endless and you can extend
81

82
00:07:18,990 --> 00:07:20,850
this in a variety ways,
82

83
00:07:20,880 --> 00:07:23,660
and as far as your imagination carries you really.
83

84
00:07:24,120 --> 00:07:28,440
But the skills are now yours for you to experiment and play with
84

85
00:07:28,440 --> 00:07:33,630
and I look forward to all of the awesome apps that you'll be building using these techniques that you've
85

86
00:07:33,630 --> 00:07:35,700
learned in this and in previous modules.
86

87
00:07:36,510 --> 00:07:38,620
So if there's anything cool that you've built,
87

88
00:07:38,640 --> 00:07:44,070
be sure to post it in the Q & A section so that myself and other students can check it out and we
88

89
00:07:44,070 --> 00:07:48,910
can congratulate you on your great work and also be your first beta testers.
89

90
00:07:48,930 --> 00:07:51,260
So I wish you all the best of luck.
90

91
00:07:51,360 --> 00:07:56,850
And in the next lesson, we're going to refactor our code to make it a little bit tidier and a little
91

92
00:07:56,850 --> 00:07:57,930
bit prettier.
92

93
00:07:57,960 --> 00:08:00,510
So if you want to watch along for that,
93

94
00:08:00,510 --> 00:08:02,520
be sure to head over to the next lesson.
94

95
00:08:02,670 --> 00:08:04,800
Otherwise, I'll see you on the next module.