Heads up! To view this whole video, sign in with your Courses account or enroll in your free 7-day trial. Sign In Enroll

Build a Chatbot with Watson APIs

Well done!

You have completed Build a Chatbot with Watson APIs!

Sign up for Treehouse Back to Library
Preview
Video Player
00:00
00:00
00:00
  1. 2x 2x
  2. 1.75x 1.75x
  3. 1.5x 1.5x
  4. 1.25x 1.25x
  5. 1.1x 1.1x
  6. 1x 1x
  7. 0.75x 0.75x
  8. 0.5x 0.5x
Use Up/Down Arrow keys to increase or decrease volume.

Trying it Out

5:10 with Ben Jakuben

The best way to understand what you can do with various Watson services is to try them out!

For starters, let's check out a website that lets you explore Watson services 0:00
right in the browser, without writing in any code. 0:03
If you search for Watson API services or click on the link in the teacher's notes. 0:06
You will come to this page that lists all the services currently available on 0:10
the Watson Platform. 0:14
So here we have some services related to conversation. 0:15
And some forward discovery and if we keep scrolling down we can see 0:19
computer vision and speech, language, and finally empathy. 0:24
As you could see there is a lot of computer empower available here. 0:29
The goal of these course, is to get you started some common and useful examples. 0:32
But there's much more available for you to use to solve different kinds of problems. 0:36
We encourage you to test a lot of these out, but 0:40
let's check out the Tone Analyzer here at the bottom. 0:42
So the Tone Analyzer service understands emotions and communication style in text. 0:47
So this means if I give it the text, I'm happy, it will detect a positive tone. 0:53
Amen, exactly how does the tone analyzer 0:57
determine human emotion from speech or text? 0:59
>> That's a great question, Ben. 1:03
We employed psychologists who use the big five personality traits. 1:05
Openness, conscientious, extraversion, agreeableness, 1:11
neuroticism, OCEAN, to be exact. 1:14
What they did is they collected lot of texts and they analyzed that text, 1:18
classified it, and they placed it within those buckets. 1:23
Each that they thought were appropriate to a particular tone that depicted 1:28
one of the big part personality traits. 1:34
From there on once Watson comes across an article, a text, 1:36
a paragraph, a noun phrase one that it hasn't' seen before. 1:41
It is able to then classify it and 1:45
come up with derivatives of those five such as joy, confidence, and 1:49
anger and hence depict the tone of the written or the spoken word. 1:53
>> It's incredible really. 2:00
And again, all this work is done behind the scenes to train these services and 2:01
make them available. 2:05
So when we're using them, 2:06
it's got all this information underneath that's ready for us. 2:07
>> Absolutely, yes. 2:11
>> So yeah, let's continue with our example here. 2:12
If we click on the view demo button, 2:15
then we come to a page that is pre-loaded with some samples. 2:17
So here we have some Tweets, Online Reviews, Email messages, etc. 2:21
And the first one in here is a series of tweets about people who are not happy 2:25
about new phone. 2:29
All right, so let's analyze this and see how it goes. 2:30
Click on the button here, all right. 2:33
So we get a document level summary, right, and it's showing us that there is some 2:35
anger, fear, and some confidence as the main tones present in this text. 2:39
And if we scroll down a little bit, we get down to the sentence level. 2:44
And we can see which sentences in the text match the different tones. 2:47
So if I click on anger, there we are, highlighted in red, 2:52
there's some angry sentences, some angry tweets about the phone. 2:55
Same thing if I click on tentative, nothing shows up. 2:58
If I try fear, got a sentence there at the bottom, about a fearful sentence. 3:02
So if we scroll back up, we can also enter our own text for analysis too. 3:07
I'm going to paste in a message I sent the teaching team recently. 3:11
So click your own text come in here and paste. 3:14
And let's see, yep, okay. 3:20
So if I analyze this. 3:21
Let's take a look, how did I do? 3:23
All right, so the tone analyzer is telling me that it detected joyful tones and 3:25
tentative tones. 3:30
If we click on each down at the sentence level we can see certain things, 3:32
the same things was detected as joyful. 3:37
If I click on tentative I can see some examples here where 3:40
we can see that the color here matches the, 3:44
it's a slightly less confident reading of this text. 3:46
So even the Tone Analyzer itself is tentative in its analysis here. 3:49
And one of the fun things that you suggested that I do is paste in some 3:55
angry text as well. 3:59
So I'm going to just here in the middle, insert some angry text and 4:00
analyze it again. 4:04
And now at the bottom, sure enough we got some angry passages. 4:06
And if I click on it, and scroll for the red one, you can not only see not only 4:10
angry but also detected some sadness which is interesting. 4:14
Two different passages, 4:18
there were different levels of confidence in how sad they were. 4:19
The Watson API documentation defines all of these interpretations. 4:23
And you can check it out using the link in teacher's notes. 4:27
You will definitely wanna refer to the docs for 4:29
each service that you wanna use in your apps. 4:32
Note that these API docs have sample code in multiple languages 4:34
right here on the right side. 4:38
This is some of the best documentation that I've ever seen. 4:40
You should plug in some things you have written, like emails or tweets and 4:43
see what the results are. 4:46
It's amazing. 4:47
>> So try some things out grab an excerpt from a sad song, 4:49
copy paste that into Tone Analyzer. 4:53
The one that I do most often is actually take a snippet of my email or 4:56
the entire email. 5:00
And I can't tell you how many times that has stopped me from sending the email 5:01
before I erase all the red and make it look blue. 5:06