Hello and welcome to Creating a Pizza Ordering Conversational Agent with Dialogflow. In this lab, you'll learn how to use Dialogflow to create a conversational interface for a pizza ordering use case. And you'll perform the following tasks, building a conversational agent for a fictional pizza shop, defining intents for a pizza ordering use case to match typical user queries, and identify entities to extract data from user queries. The first task is to enable the data flow API. For security purposes, many Google cloud services are disabled by default, and this is referred to as the principle of least privilege. So you must enable the specific APIs needed by this lab before beginning it. In the GCP console, on the Navigation menu, click APIs & Services. Then click ENABLE APIS AND SERVICES. And search for Dialogflow. Then select Dialogflow API, And click Enable, if it's not already enabled. Now you're ready to actually start using Dialogflow. Dialogflow requires a Google account to sign in. And you'll use the quick lab account to sign in to complete this lab. Start with Dialogflow console at the website https ://console.dialogflow.com. Click on Google to sign in. Allow Dialogflow to access your Google account. Lastly, you'll be taken to Dialogflow's Terms of Service, which you'll need to accept in order to use Dialogflow. Let's create our first Dialogflow agent. To create a Dialogflow agent, click CREATE AGENT in either the left menu or the CREATE AGENT button on the right of the screen. Enter pizzaBot for your agent's name, English as the default language, and set the default time zone to your time zone. Ensure you're using the same project that you see in your GCP Console. Then click the Create button. You should now see the Dialogflow console. On the left is the menu panel. If you are working on a smaller screen and the menu is hidden, click on the menu button on the upper left corner. The settings button takes you to the current agents settings. The middle of the page will show you the list of intents for the agent. Generally, an intent represents one dialog term within the conversation. By default, Dialogflow agents start with two intents. Your agent matches default fall back intend when it doesn't understand what users say. The default welcome intent greets your users. These can be altered to customize the experience. Later in this lab you'll create your own intents to match user intentions, like ordering a pizza for example. On the right is the Dialogflow simulator. This lets you try out your agent by speaking or by typing messages. Let's type two messages to see if we can activate both different intents. Hello, an the agent responds, hi, how are you doing? And notice that activates default welcome intent. And if I now write, can you make me a milk shake? The agent has no idea how to interpret this and so it falls back to the default fall back intent and responds, say that one more time. The default welcome intent is preconfigured to understand basic greetings and to reply with a generic response. Let's replace the generic greeting with something that let's our users know who we are. First, click on the Default Welcome Intent. In the responses section of the Default Welcome Intent, remove all of the predefined responses like good day, greetings, and others. And add the following two responses. Welcome to our store, I can assist you store hours or place an order. Which would you like? And Hello there. I can assist you store hours or place an order. How may I help you? And then click Save at the top. Now let's try it out by saying hello to our agent and seeing how it responds. Notice the default response is one that we just added in. In the same way we just did for the default welcome intent, you may customize the responses for the default fallback intent. Additionally, besides the ones provided by default, we'll need to create custom intents that will answer specific queries for our pizza use case. So let's create our own intent. Dialogflow uses intents to categorize a user's intentions. Intents have training phases which are examples of what a user might say to your agent. For example, someone wanting to know what hours the pizza shop is open might ask, when do you open, what are your hours, or how late can I come in? All these queries are unique but have the same intention, to get information on when the pizza shop is open. To cover this query, create an opening hours intent. Click on the + next to Intents in the left menu. Add the name opening hours in the Intent name text field. In the Training phrases section, click on ADD TRAINING PHRASES and enter the following pressing Enter after each entry. What time do you open? Are you open right now? Are you open today? And when do you close? In the Responses section under text responses, enter the following Response in the text field. We are open from 10 AM to 9 PM everyday. Can I help you place an order? You may also notice the message Agent training started or Agent training completed on the bottom right of the screen. Once training is done, try out using the simulator on the right by asking the question, which days are you open? And the response is, we are open from 10 AM to 9 PM everyday. Can I help you place an order? Let's start extracting data with entities. In this section we'll go over how to extract data from the user's query to let them place their pizza orders. For example, agent will not only need to know that a user wants to place an order for pizza, but it is also important to know the details or parameters like size, toppings, and the day and time the user wants to place the order for. For this purpose, you'll create another content called order pizza, and the training phrases you provide, identify these important parameters. Parameters are important to and relevant words or phrases in a users query then extracted, so your agent can provide a proper response. Create a new intent by clicking on the plus add next to content in the left menu. Name the intent order.pizza. Add the following as training phrases like can I order a cheese pizza for pickup today at noon, etc. I've already done so, and as you can see Dialogflow has automatically highlighted the entities that it knows about and then created them in the action and parameters section. For example, the first two have both date and time in them. And you'll notice in the action and parameters section that there's a date and time parameter. These values occurs in the action and parameter section in a table that describe all the different characteristics of the parameters including their names, the entity, as well as the value whether or not it's a list, and whether it's optional. The value of a parameter can be used in your responses. In this case, you can use date and time in your response. And it will be replaced with the date and time specified in the query to your agent. In the Responses section, add the following response and click the Save button. Great, we've placed your order for date at time, see you then. Now query your agent with, I want to order a cheese pizza for pick up today at 5 PM, and see what happens. And there you have it. It used the order pizza intent, and it recognized both the date and time, and included it in the response. What we've done so far is great, but let's not forget that we still need to extract other parameter values, like pizza toppings and size, for which the entities don't exist yet. In this next section, we'll create them. Click on the plus next to Entities in the left menu. Add the name size into the Entity name text field. And make sure that the Define synonyms check box is checked. Next, add the three reference values and one synonym for each. So we're going to add small as a reference value with the synonym small, medium with the synonym medium, and then large with the synonym large. And then click Save when you're done. In the same way we're going to create a whole bunch of more synonyms for each of the pizza toppings that we intend to recognize. Some of them will have more synonyms than others. When you're finished, it should look something like the table you see before you. So we've created a whole bunch of entities, but what we need to do now is map the parameters and the intents that we receive to those entities. To do so, first click on the Intents in the left menu and then click on the order pizza intent. Look at the first training phrase and highlight the word that is either the topping or the size of the pizza being ordered. In the case of cheese, it should map to the pizza topping entity. And if you don't see it in the list, start typing it out in the text box on top and it should be filtered. Now you'll notice that for this particular training phrase there are three entities identified. In the same way, look at the other training phrases and for each word that describes the size or the topping highlighted and associated either with the pit size or pizza topping entities respectively. Note that some phrases have multiple toppings and so highlight each and associate each with a pizza topping entity. When you're finished, it should look like so. Once you've done this for all training phrases, you can see all parameters and their entities in the actions and parameters box. You can now modify the responses section to include the additional new parameters in your default text response. To do so, simply click on the text response. And then whenever you intend to insert the parameter, type the money sign, and it should auto complete with a list of optional parameters. In this case I've modified it to include both the size and the pizza topping. At this point, our Dialogflow chat bot is capable of taking orders. However, it doesn't require the user to actually enter any of the parameters that we've specified. So if I enter the order, order pizza, in this case it will say, we've placed your order. And that's in part because I've added this text response here, we have placed your order that doesn't actually use any of the parameters, and it thinks it's a valid option. What we need to do is somehow tell the chat bot that these parameters are required in order to proceed. To do that, we're going to use something called slot filling. To do slot filling, we need to scroll to the actions and parameters box and check the check box to the left of each parameter to make it required. As you check the checkbox you will see a new column added for each parameter called prompts. These are the prompts or questions presented to a user to get the desired parameter value filled. For example, if the user says I want a chicken pizza, the size parameter could ask, what's size pizza do you want? The next task then is for each parameter to define a prompt that indicates the user that they must enter the value for that parameter. So for example, for date, you could say something like, what day do you want to pick up your order? And so forth. I have now entered prompts for every single one of our parameters and saved the intent, which triggered the agent to retrain. If our efforts were successful, if we try an order a pizza and the agent does not receive values for all of our required parameters, it should prompt us with one of the prompts that I specified. So when I type order pizza, as I have done, it prompts us with a prompt for the day. So I can give it a day. Let's see what happens. So now that it has a day, it's asking us about the time. And now that it has the time, it wants to know the size. And finally, it wants to know the toppings. Once it has all the required parameters and does indeed give us the default response with the parameters imbedded in it. Note as well that the action section of the this dialog window contains the values of this parameters. And that's it for this lab. But you've now created an agent that will be able to answer user queries around placing an order for pizza. The next step, which isn't covered in this lab, would be to build functionality to actually be able to place the pizza order into a database and connect it to other downstream systems. And you can do this with fulfillment.
