01_LexBotInformational

Module 1: Build a chatbot in Amazon Lex and handle informational queries

In this module you will build a chatbot in Lex and enable it to answer customer's questions on the offerings of your company.

At completion of this module you will be able to test out your chatbot in the Lex console by asking questions such as "Tell me about international plans in Germany."

Informational queries and chatbots

The first intent you will configure your bot to understand and fulfill will allow your customers to query phone plan options when they travel abroad. This serves an example for a common use case for chatbots: asking information.

There are two ways you can implement an informational intent: you can make it general or personalize it based on the customer. The latter one requires your bot to be able to identity the customer the bot is interacting with.

For example, if you build a chatbot that can handle checking the flight status, the general implementation can handle questions like "what's the status for flight 123?"; whereas if you can identify the customer asking the question, the bot can give answers to questions to "what's the status for my flight today?"

In later modules, you will try out ways for identifying the customer in a Lex bot conversation. For this module, we will stick with a general information query: what are the available phone plan options for a given country the customer is traveling to? The bot will give the same answer for anybody that asks the same question.

Implementation Instructions

Each of the following sections provide an implementation overview and detailed, step-by-step instructions. The overview should provide enough context for you to complete the implementation if you're already familiar with the AWS Management Console or you want to explore the services yourself without following a walkthrough.

Prepare resources

In this step, we will use a CloudFormation template to provision the AWS Lambda functions, DynamoDB tables, resources we will need in later steps of the workshop.

Region	Region Code	Launch
US East (N. Virginia)	us-east-1

CloudFormation Launch Instructions (expand for details)

1. Click the Launch Stack link above.

2. Click Next on the Select Template page.

3. Click Next on the Specify Details page.

4. On the Options page, leave all the defaults and click Next.

5. On the Review page, check the boxes to acknowledge that CloudFormation will create IAM resources.

   

6. Click Create Change Set.

   Note the CloudFormation template we've provided is written using AWS SAM (AWS Serverless Application Model). SAM simplifies how to define functions, APIs, etc. for serverless applications, as well as some features for these services like environment variables. When deploying SAM templates in CloudFormation template, a transform step is required to convert the SAM template into standard CloudFormation, thus you must click the Create Change Set button to make the transform happen.

7. Wait for the change set to finish computing changes and click Execute

8. Let the CloudFormation launch resources in the background, you don't need to wait for it to finish before proceeding to the next step.

1A: Create a Lex bot

Use the Lex Console to create a bot named InternationalPlan.

Step-by-step instructions (expand for details)

1. Go to the Lex Console.

2. If it's your first time creating Lex chatbots, click Get Started.

   If you have created Lex bots before, click Create under the Bots tab.

3. Pick Custom bot (create your own).

4. Fill in the form:

   For Bot name, use InternationalPlan

   For Output voice, pick Joanna

   For Session timeout, use 10 minutes

   This is how long your session context will be maintained so your user don't have to verify their identity again if they are interacting with the same bot and device in that time period.

   For COPPA, pick No.

5. Click Create

1B: Create the first intent

Create a new intent ListInternationalPlans in the Lex bot

Step-by-step instructions (expand for details)

1. In the InternationalPlan Lex bot you just created, click +Create Intent

2. Pick Create new intent

3. Give the intent a name, ListInternationalPlans, then click Add

1C: Configure Slots

Slots are parameters you can define to capture inputs from your customer. In this example, the input parameter the bot need in order to fulfill the informational query is which country the customer is traveling to.

Each slot has a type. You can create your custom slot types or use built-in slot types. Check the list of Built-in Slot Types

Because there is a built-in slot for country names, we will leverage it for this intent.

Configure a slot Country with built-in type AMAZON.Country in the intent.

Step-by-step instructions (expand for details)

1. In the Slots section of the ListInternationalPlans intent, fill in Country for the slot Name

2. Select AMAZON.Country for Slot type

3. For Prompt, put in Which country are you traveling to?

4. Click the (+) sign to add the slot

   

1D: Configure sample utterances

By providing sample utterances for a given intent, you can teach Amazon Lex different ways a user might convey an intent.

Add the following sample utterances to the intent:

• I'm traveling to ​{Country}​
• to ​{Country}​
• List international plans
• List international plans for {Country}
• List travel plans available
• Tell me about travel plans in ​{Country}​
• I want to know about travel plans in ​{Country}​
• What plans are there for ​{Country}​
• What international plans do you have

Note that you don't need to list exhaustively every possible way of saying the same intent, just a few examples so the Amazon Lex deep learning algorithms can "learn".

However, if during testing you identified some additional ways to express the intent and Lex doesn't understand it, you can add that as a sample utterance to improve the Lex bot.

At this stage, we haven't configured the backend logic to look up actual plan options the user asks for. But we can test how our Lex bot can understand customer requests before we integrate the backend.

Save the intent, build and test the bot in the Lex Console.

Step-by-step instructions (expand for details)

1. Click Save Intent to save the intent configuration

2. Click Build at the top right of the page to build the bot

3. Once the build completes, use the Test Bot window to test different ways customer may ask about international plans for the countries they are traveling to. Verify that the bot is able to detect the intent.

   In the below example, the user utterance contains the slot value, which Lex was able to detect:

   

   In this below example, the user didn't tell the country he/she is inquiring about, Lex will use the prompt we configured for this slot to get this info from the user:

   

1E: Fulfill the query with AWS Lambda

Now we have defined the conversational interface, we need to configure the backend logic to fulfill the customer query using AWS Lambda.

The Lambda function that can respond to the international plan customer request is already deployed by CloudFormation in the setup step. (It does so by querying a DynamoDB table pre-populated with fake data, also launched as part of the preparation CloudFormation)

Expand here for instructions to check the DynamoDB table content on international plan catalogue

1. Go to the DynamoDB console

2. Select the table name starting with lex-workshop-TravelPlanCatalog

   

3. You should see a list of pre-populated fake international plans. (Additional columns such as price per text are provided so you can use them to extend the bot. e.g. add a GetPlanDetails intent)

   

Now we are ready to configure Lex to send the detected intent and slot values from the user utterance to the lex-workshop-LexBotHandler Lambda function.

Step-by-step instructions (expand for details)

1. In the Fulfillment section of the intent, choose AWS Lambda function and use the selector to pick the lex-workshop-LexBotHandler function

   

   There are a handful of other Lambda functions the CloudFormation template created and that they all begin with lex-workshop, so be sure to select the right one.

2. Click OK to give Lex permission to invoke the Lambda function.

3. Save the intent by clicking Save intent

4. Build the bot again by clicking Build

5. Test the bot

   

   The plan data is randomly generated and loaded into a dynamoDB table by the CloudFormation. It might not always make economic sense.

6. Feel free to test the voice interaction in the Console as well.

1F: Check execution logs for Lambda

It's also valuable to understand what data is being passed to your Lambda function. Take a look at the lex-workshop-LexBotHandler function's CloudWatch Logs

Step-by-step instructions (expand for details)

1. Go to the Lambda console

2. Find the lex-workshop-LexBotHandler function and click on it

3. Go to the Monitoring tab

4. Click View logs in CloudWatch

5. Click on the latest log stream

6. Find the log line that logs the input into the lambda function:

   

7. Observe the fields being passed from Lex to Lambda: userId, bot, inputTranscript, name of the intent, and slots identified. See documentation here on detailed explanation of all available fields.

   A note on the userId field:

   Think of it as a session identifier used to distinguish conversations or threads. If you are building integration using Lex's API directly, see documentation here on deciding what value to use for the user ID field. For natively supported messaging platforms, the userID is filled for you by the integration (e.g. the user's phone number is used as userId in the case of Twilio SMS.)

Next module

After you have verified your bot is answering questions correctly and checked out the lambda execution logs, move onto the next module: Handle customer requests to subscribe to services