Please enable JavaScript.

Coggle requires JavaScript to display documents.

Creating your chatbot - Coggle Diagram

- - - - visit https://dialogflow.cloud.google.com/(opens in a new tab).
    - - If you're logging in Dialogflow for the first time, you should see a blue "CREATE AGENT" button in the middle. Alternatively, you can click on "+ Create Agent" near the top left.
    - - We will give our agent a name - CarreyTheCarAgent. Here, you can give it any name you want, of course, so long it is easily understandable. When you have more agents in future, you want to be sure of the purpose of each agent.
        
        Ensure Carrey understands English conversation, and she is in the correct timezone of GMT+8.
        
        Click on the blue "CREATE" button once done.
- - - - The "Default Welcome Intent" and "Default Fallback Intent" are already there. Let's click on "Default Welcome Intent" and see what it says!
    - - You now enter the "Default Welcome Intent". All changes made on this page will affect this intent.
        
        Scroll down and you will see a list of Training Phrases. These are what the user might type to the chatbot. For example, the user could say "heya" or "greetings" etc.
        
        Remember, training phrases are what the customer might type to the chatbot.
    - - Further down the "Default Welcome Intent" page, you will find there is a section for you to define parameters (and entities). We will discuss this later.
        
        Scroll on down and you will see the "Responses" section. Based on what the end-user says, the agent will match to see if it is similar to the Training Phrases above, and if so, it's a match and the agent will reply any one of the text response.
        
        You can also add a response variant -- give it a try. Type "Hello, I am Carrey. How can I help you?" into the text response variant.
        
        Click "SAVE" at the top of the page. Your agent will begin to train itself, and the setting icon at the top left will be spinning. This will take about 10 minutes, so grab a coffee!
    - - Your Settings icon at the top left should stop spinning before you perform this step. Alternatively, skip this step and come back to it later.
        
        You can try out the Test Console on the right side of the screen. Type "hi hi" [hi space hi] into the console and see how your chatbot would respond.
        
        Notice how "hi hi" is not an exact match with any of the Training Phrases, but the system still correct identifies the intent as the "Default Welcome Intent". Also, the response is any one from the list.
        
        Try to type a few different ways that you would say hi to a chatbot, and see the response. At this time, you can also add more Training Phrases or/and Responses to this "Default Welcome Intent" if needed.
    - - Each time you add training phrases or responses, remember to click "SAVE" to apply all changes you made to the "Default Welcome Intent".
        Now, click on "Intents" on the left panel to go back to the list of intent. Review "Default Fallback Intent", which is what happens when your chatbot cannot understand your user. Add in more Response Variants if needed.
    - - We will build our chatbot such that the agent can handle different intentions of customers. To do so, click on "Intents" on the left panel, then click on "CREATE INTENT".
    - - For this intent, we also need to think of different ways which customer will ask ABC Automotive Workshop about the types of services it provides. Add the following to training phrases, and think of even more ways customers might say this.
    - - Create more intents as outlined in this document(opens in a new tab). Be sure to input the respective training phrases and responses for each intent.
        When complete, your list of intents should look like the image above.
- - - - Click on "Entities", then click on "CREATE ENTITY".
    - - In this entity, we will only type in two possible values - "car" and "truck", since we assume ABC Automotive Workshop only handles these two types of vehicles.
    - - In the "Rental" intent go to Actions and Parameters. We will add in the entity we created.
    - - For all the training phrases under this intent, highlight the words "car" and "truck" and click VehType:VehType so that the agent will know they are defined entities.
    - - Because we set the entity VehType as "Required", when the user does not mention whether he wishes to rent a "car" or "truck", your chatbot will prompt for VehType.
    - - With the use of an entity placeholder, you can use the entity value specified by the user. Under the intent "Rental", scroll down and modify the Response to "We've got it! Visit our car workshop to rent a $VehType."
    - - Add another parameter as shown in the image above, and modify the response by adding in "on $date". Remember to click "SAVE".
- - - - On the left panel, click on "Small Talk" and enable small talk. By doing so, we can teach our agent how to reply to phrases like "thank you" and "you look great today" - just in case anyone decides to praise our bot.
    - - Dialogflow provides a validation feature to help us create high-quality agents. Agent validation results are available automatically whenever agent training is complete. These enable us to know if there are any issues with our agents.
        To enable agent validation, click on the "Setting" icon, go to "ML Settings", then "Enable" Agent Validation and "TRAIN" your agent. Remember to click "SAVE".
    - - On the left panel, click on "Validation". Here, you can see any Info, Warning and Error that exists in your agent. Expand on the issues to read more on how you can improve your agent.
    - - On the left panel, click on "Training". Here, you can see all the conversations made with your chatbot, and manually train your chatbot better. (Note: All conversations can also be accessed from the History button.)
    - - For each end-user expression, if the agent correctly matches the intent, you can click on the tick icon. This allows the agent to definitively know it matches the intent correctly when it encounters the same phrase again in future.
- - - - Mouseover the "Servicing" intent, and click on "Add follow-up intent" > custom
    - - You will see a new "Servicing - custom" intent. Click on it and configure the following:
        
        Intent Name: Servicing - Car
        
        Training Phrase: car
        
        Response: For cars, it is a flat $300 for servicing. All checks and necessary parts replacement are included in this price.
        
        Click "SAVE".
        
        Now, try to type "i wanna do servicing" in the Test Console, and when prompted for the vehicle type, type "car" into the Test Console.
    - - Create another follow-up intent under "Servicing" using the instructions below, and when done, your list of intents should look like the above.
        Mouseover the "Servicing" intent again, and click on "Add follow-up intent" > custom. For this follow-up intent, configure
- - - - Botfather is a bot that can help us to create chatbots on Telegram. Search for botfather in Telegram.
    - - Hit the "START" button to begin creating our chatbot, and we will be guided through step by step.
    - - Type /newbot into the message bar.
    - - Public users can search for your chatbot using this name, so it should be a unique name for you or your business.
    - - Earlier, we gave our chatbot a name, and here, we will give it a username. The username must end with _bot. Note that your chatbot cannot have the username abccarworkshop1234_bot, simply because I have taken up this username!
    - - You will see a token that is generated - this is a very long string of text and numbers. Keep this token private and secure, as anyone with the token can control your chatbot!
    - - In Dialogflow, we will now need to integrate our Dialogflow agent to our Telegram chatbot. Click on "Integrations" on the left panel, and scroll down to select "Telegram".
    - - Enter the token generated by BotFather and press "START", then "CLOSE". Your bot is now live on Telegram!
    - - Search Telegram for your chatbot name (or search for the username). Give it a go and see how it works!