(Music) This module has two labs. One, in which we’ll create the basic chatbot for our student advisor, and one in which we’ll add Watson Discovery responses to the chatbot. In this video, I’ll show you what you’ll do in the first of these two labs. As always, do not worry about following the steps right now. Just focus on the video, as you’ll perform the same steps while taking the actual labs. We’ll start off by launching our Watson Assistant instance. Here, we’ll create a new assistant for our student advisor. We’ll give it the name Student Advisor Chatbot and we’ll leave the preview link enabled so that we can quickly share the chatbot for testing purposes with friends and colleagues. Now that we have an assistant, we’ll add a dialog skill to it. We’ll simply call it Student Advisor, and since our chatbot will be in English, we’ll ensure that the English language is selected. With the Dialog Skill created, we’ll select the Content Catalog. Here we’ll see a list of pre-made intents that are common to most chatbots and for various industries. We’ll add General to our skill, so that our chatbot is capable of handling chit chat. A series of intents will be added to our list of intents. They are all prefixed by General, making them easy to distinguish from the intents that we’ll create ourselves. On that note, let’s create an intent. We’ll call it Enrollment_Cost and add a few examples of queries our students might have in relation to the cost of our courses. Next, we’ll create an intent to handle people who ask questions about Coursera’s professional certificates. As always, we’ll add a few examples of the questions users might have. For example, “Which data science certificate?”. With the intents taken care of, we can now focus on adding some entities to capture specifics about the user input. We’ll create a professional certificate entity. After all, there are only so many professional certificates available so it’s okay for us to detect them all through an entity. We’ll add a value to the entity for each professional certificate, as well as some synonyms, so that data analysis and data engineering, for instance, are also detected as examples of people inquiring about the data science certificate. By the way, this very course is part of a new AI professional certificate that will be live by the time you watch this video. Make sure you check it out. We have our chit chat intents, an intent for questions about costs, and one about professional certificates along with an entity to distinguish between the various certificates available. What we don’t have yet is an intent for course recommendations. Let’s create that intent. We’ll call it Course_Recommendation and add some examples, such as “I want to learn Python” or “Recommend me a course on nutrition”. Now that we are done with our basic intents and entities, it’s time to create the dialog. The Welcome and Anything else node will created by default for us. We’ll edit the Welcome node to improve the chatbot prompt. In the response, we’ll add, “Hello. I’m a Student Advisor Chatbot. I can help you with your questions about our site and give your course recommendations.” Now we’ll need to handle at least some basic chit chat interactions, and to keep the chatbot nice and organized, we’ll add a folder just for chit chat nodes. This folder is just a container to keep us more organized, so we don’t need to specify a condition. Next, we’ll add a node to the folder. We’ll start off with a Greetings node which uses our General_Greetings intent. We’ll repeat the process for another child node. This one is for thank yous from the user. We’ll use the General_Positive_Feedback intent as a condition for the node, and provide some typical responses to be selected randomly each time by the chatbot. For example, you’re very welcome or no problem. Finally, we’ll add out third child node to the Chitchat folder. This node will handle goodbyes. The condition will be the intent General_Ending, and the responses will be a series of goodbye variations, such as ”See you. Have a great day!” The chatbot can now greet the user, handle basic chit chat, and thanks to the Anything Else node it can inform the user when it fails to understand them. It’s not much but our chatbot is clearly starting to take shape. After collapsing the Chitchat folder, we’ll be able to select the Welcome node to add a node below it. This node will handle enrollment cost enquiries and will therefore have an Enrollment_Cost intent as its condition and an appropriate explanation in its response. and will therefore have an Enrollment_Cost intent as its condition and an appropriate explanation in its response. Next, we’ll select the Enrollment Cost node we just created to create another node below it. This node, you guessed it, will handle professional certificate recommendations and requests. The condition for the node to be executed is the Professional_Certificate_Recommendation intent. The response will depend on the type of professional certificate the user asked us about, so we’ll need to customize the node to switch from a regular text response to a conditional multiple response With multiple responses enabled, we’ll be able to provide a specific response for each professional certificate entity value available. Finally, we’ll have a response whose condition is true and it will act as our fallback case. This response will provide a link to the page on the Coursera site which lists all professional certificates. We’ll repeat the process for our third and last domain-specific node. Namely, our Courses node. For this node, we’ll set the condition to the Course_Recommendation intent and leave the response empty for now. In fact, the Coursera catalog has just too many courses to hardcode in responses for each of them. Instead, we’ll programmatically access to a list of relevant courses by leveraging Discovery in Lab 4, later in this module. As a quick sanity test, we can launch the Try It Out panel and verify that the chatbot responds correctly to a sample query such as “can you recommend a professional certificate?”. Now, it’s your turn. In the next lab, you’ll repeat these steps yourself. By the end of the lab, you’ll have the same barebone chatbot as the one you saw in this video. Good luck. (Music)
