Creating an Intent4:25 with Ben Jakuben
An intent communicates a user's goal, something they want to achieve. We will start building our Conversation service by defining our first intent.
We will start building our conversation service by defining our first intent. 0:00 An intent is a group of examples of things to use or might say to do something. 0:05 They're communicating their goals, something they want to achieve. 0:11 Creating an intent is easy. 0:16 You start by thinking of something your users will want to do. 0:18 Then you went to the multiple examples of how they would say or type that goal. 0:23 >> In our current chaoter, our user would want to turn things up, for example, 0:28 they might think something like, I want to turn on the radio and 0:32 we want Watson to learn how to handle this. 0:35 So what is the goal here? 0:38 They want to turn on the radio. 0:39 Now, for intents, 0:41 we only care about what they want to do, which in this case is turn something on. 0:42 We don't care yet whether they're talking about the radio, lights, the heater or 0:47 whatever, we just need Watson to realize that the intent is to turn something on. 0:51 Okay, so, if we're in the app right now, where it's not doing anything, and 0:57 we type, turn on the radio. 1:00 As expected, nothing happens. 1:03 But, over here on the right, we have this section called Watson understands. 1:05 In this, is the JSON response currently being returned 1:10 by our conversation service. 1:13 And, what I want to highlight is here. 1:15 Notice that there is an intense array here and it is empty. 1:17 So that's what we're going to build. 1:21 So Arman, how should we get started? 1:23 >> Excellent and wonderful. 1:26 Let's go to our conversation service and 1:27 we're going to open the treehouse workspace. 1:30 An intention may be to turn something on or 1:34 to turn something off, we don't know what yet. 1:37 So why don't we start with turn underscore on? 1:41 No spaces in this. 1:45 >> Okay. >> And the pound sign, and later on you'll 1:47 note the at sign with entities, is not a programmatical construct. 1:50 It's merely here to help distinguish these two 1:55 easily as we move- >> Okay. 1:59 Cool. >> Into the dialogue. 2:01 >> Makes sense. 2:03 >> Well, let's enter an example. 2:03 For example, power on. 2:06 >> Okay. 2:10 >> Turn on. 2:13 Switch on. 2:15 Play. 2:20 Start. 2:23 >> Good, good, sometimes I'll say flip on, like flip on the radio. 2:25 >> Absolutely, absolutely. 2:28 >> So now I noticed that we're not adding the entity yet, 2:30 like we're not saying, power on the radio, or turn on the wipers. 2:33 That's because if I'm understanding the intent just wants to focus on 2:36 again that we are turning something on. 2:40 Is that correct? 2:41 >> You're precisely right, Ben, absolutely. 2:42 In here we are defining our intention and as we become more gradual, we will then in 2:44 the intense, identify what is it that we want to turn on or turn off and similar. 2:50 >> Okay, so having a wide variety of examples here is really helpful. 2:56 The more we can add to the intent, the better it will help Watson 2:59 learn how to recognize that the user wants to turn something on. 3:03 So we don't need to list every possibility, but 3:06 we do want to list at least five examples for each intent that we create. 3:08 And we can always edit or add these examples later. 3:12 >> Absolutely true. 3:15 And bear in mind that natural language classifier, which is 3:16 the engine running intents and entities, already comes trained with Wikipedia. 3:21 All we need to do is just give it a push in the right direction. 3:27 >> Very neat, yeah. 3:30 >> We should also point out that you can import data for intents and entities, 3:32 which is helpful if you have a list already or generated data somewhere. 3:36 We'll see how to do this a bit later. 3:42 >> Okay, so I guess I can click Done now ,right? 3:45 This one is done. >> Yes. 3:47 >> Good, and now we can go back to our example, and I'm gonna refresh. 3:48 And now let's try again one of our examples, turn on the radio. 3:56 And over here in the Watson Understands section, now the intents is populated 4:03 with an array and it matched the turn on intent that was just created, and 4:08 we gave it with a very high confidence score here of 0.95, 4:13 which a score of 1 is like it's perfectly confident. 4:17 So Watson at this point is very confident that we want to turn something on. 4:20
You need to sign up for Treehouse in order to download course files.Sign up