EIE1005 Fundamental AI and Data Analytics Workshop 3R

Java Python EIE1005 Fundamental AI and Data Analytics

Workshop 3: Creating Chatbot using DialogFlow

A. Objectives and Outcomes

After finishing this lab, you should be able to

•    understand the practical aspects of conversational AI

•    know the basic components in DialogFlow

•    master the fundamental process of constructing chatbots using DialogFlow

B. Assessment Criteria

•    Ability to design and build chatbots using DialogFlow

•    Ability to produce a working chatbot

•    Ability to create a chatbot that is slightly different from the example shown in the lab procedure

•    Clarity of the report

C. Submission

• In-class demonstration

•    Capture the screen and put the pictures in your report.

• Convert your report to PDF.

•    Submit the PDF file to Blackboard before the deadline specified in Blackboard.

D. Getting Start

1.    We will use Google DialogFlow. You need a Google account to use DialogFlow. If you do not have a Google account, visit https://support.google.com/mail/answer/56256?hl=ento create one. You may access DialogFlow through any browser, and no installation is required. It is free!

2. Sign in DialogFlow by visitinghttps://dialogflow.cloud.google.com/

3.    To start with a new Chatbot, you need to create an agent. Click “Create Agent” on the left panel.

Give a name to the agent (e.g., BikeShop) and click “Create”. You should see something like the following.

4. Enable text-to-speech by clicking the setting symbol on the left panel. Select the “Speech”

tab and “Enable Automatic Text-to-Speech”. Then, click “Save” .

5. Customize the Default Welcome Intent:

a.   Click on the Default Welcome Intent

b.   Navigate to the Responses section

c.   In the Text or SSML response table, delete all the default responses.

d.   In the Text or SSML response table, copy and paste the following response: “Welcome.

I can tell you about the shop hours, or I can set up an appointment. Which would you like?”

e.   Click the “Save” button

f.    Test your Welcome Intent by putting “ Hi” in the “Try  it now” edit box on the right panel.

6. Customize the Default Fallback Intent.

a.   Click on the Default Fallback Intent.

b.   Navigate to the Responses section.

c.   In the Text or SSML response table, delete all the default responses.

d.   In the Text or SSML response table, copy and paste the following response: “Sorry, did you want to hear our hours, or set up an appointment?”

e.   Click ”Save”

f.    Try something irrelevant on the “Try it now” edit box on the right.

7. Add some more welcome messages to the “Text or SSML Response” entry. Your welcome messages will be randomly picked by DialogFlow when the chatbot operates.

8. Create  Custom Intent. We want the  agent to inform.  customers about the opening hours and schedule appointments for customers. Create a new intent named Hours by clicking the + icon. Then, click “Add Training Phrases” . In the Training phrases section of the Hours intent, enter the following training phrase: “When are you open?”.  In the Responses section, add “We're open from 9 AM to 6 PM dai 写EIE1005 Fundamental AI and Data Analytics Workshop 3R every day. Is there anything else I can do for you?”

9. Adding more training phrases. In natural conversations, we almost always have different ways of forming phrases that mean the same thing. The more training phrases you provide for intent, the more likely that user utterances are correctly matched with the intent.  Add the following training phrases:

a.   Are you open today?

b.   How late are you open on weekends?

c.   When do you close?

d.   What time do you open tomorrow morning?

e.   Are you open now?

f.    Business hours.

g.   How early can I drop in?

h.   Tell me your opening hours.

i.    What are your hours?

j.    How late can I come in?

Click “Save” and try something similar to the above training phrases in the “Try it now” edit box.

10.  To allow customers to make appointments, an Intent should be able to collect the date and time information from the customers. To this end, we may create a new Intent that accepts input parameters. The entities and parameters of an Intent allow the agent to extract targeted information from user utterances and convert it to a set of parameters, which then can be processed by other functions or systems to perform. various tasks.

Create an intent named “Make Appointment” . In the Training phrases section, add the training phrase: “Can I schedule service for 4 PM tomorrow?” In the Action and parameters table, verify that the system entity @sys.date-time has appeared.

In the “Text response table”, add the response phrase: “Got it. I have your appointment scheduled on $date-time. See you soon. Good-bye.” Test your new intent by entering “Can I schedule service for 5 PM tomorrow?”

11. Intent with more interaction. Sometimes users may only provide the date without the time. We

could request users to provide the time in the next dialog by setting the entity as “Required”.

Add a training phrase for date parameter:

Add + New parameter for time (you may need to add a training phrase for time also):

Add Response for date and time:

Try the input “May I come on Tuesday?”, followed by 10:00 AM.

12. Integration with Dialogflow Messenger. You may make your chatbot accessible on the browser.

Click “Integrations” and select “Dialogflow Messenger” . Then “enable” and “Try it now” .

In-class demonstration:

Demonstrate the Dialogflow Messenger to our TAs/Technicians and sign on the mark sheet in class.  Marks will only be given to those who signed the mark sheet.

E. Design Your Own Chatbot

13.     This part is more challenging. If you are happy with getting 70% marks in this workshop, skip this part. You should create a chatbot for a company in your discipline. For example, if you are optometry students, you may design a chatbot for an optical and contact lens shop. If you are from Design School, you may design a chatbot for a design house or fashion shop. Your marks will depend on your innovation and functionality of your chatbot. You should show in your report that your chatbot works as expected. You may refer to the following links to see more advanced features of DialogFlow. Because it is a 4-hour workshop, you do not need to implement a full- fledged chatbot.

For the report format, I encourage you to design your style. However, the report should contain all the results of the workshop