NameMatchBot

Bot for matching names against a database of names using fuzzy matching techniques. The Bot's Lambda function uses a multiple step dialog strategy for eliciting the name from the user. First, it prompts for spoken name. If it cannot find an exact match then it attempts to use a Soundex code to match the name. If unsuccessful it then re-prompts the user to spell their name. The bot then uses a Levenshtein algorithm with a threshold of two letter mismatch to find matches.

AWS Resources

The following shows the resources created for Lex, Lambda, S3 and DynamoDB along with their relationships.

Installation Instructions

To recreate the bot and names database used for the blog post you will need 2 unique buckets in your AWS Account. One is used for the Lambda Layer which imports the PyPhonetics library (unique-bucket-1) and the second (unique-bucket-2) is used to populate the Names database.

1. Create the PyPhonetics Layer for AWS Lambda from the pip install.

   • Create a folder named 'python' in your current directory.

     mkdir python
     

   • Install PyPhonetics into the python folder.

     pip install --target=./python PyPhonetics
     

   • Zip the folder into a file named PyPhoneticsLayer.zip

     zip -r PyPhoneticsLayer.zip python
     

2. Upload the Lambda Layer for PyPhonetics library to S3.

   • Create s3 bucket with a unique name, replace unique-bucket-name-1 with a unique name of your choice. This bucket will be used to store our Python dependency for the Bot Lambda.

     aws s3 mb s3://<replace with unique-bucket-name-1> --region <region>
     

   • Upload Lambda Layer to your S3 bucket from previous step.

     aws s3 cp PyPhoneticsLayer.zip s3://<replace with unique-bucket-name-1>
     

   • Upload Cloud Formation Template to the S3 bucket from step a.

     aws s3 cp BlogPost-Infra-CFN.json s3://<replace with unique-bucket-name-1>
     

3. Execute the CloudFormation script. Replace unique-bucket-name-1 with the name from the previous step and use a second unique bucket name for unique-bucket-name-2. This second bucket will be used to upload a CSV file of first and last names to populate our DynamoDb table of names.

   aws cloudformation create-stack --stack-name namesblogpoststack \
       --template-url https://<replace with unique-bucket-name-1>.s3.amazonaws.com/BlogPost-Infra-CFN.json \
       --parameters    ParameterKey=PyPhoneticsLayerBucketName,ParameterValue=<replace with unique-bucket-name-1> \
                       ParameterKey=BucketName,ParameterValue=<replace with unique-bucket-name-2>  \
       --capabilities CAPABILITY_NAMED_IAM \
       --region <region>  
   

4. Verify the CloudFormation script has completed successfully. The cloudformation stack will take several minutes to finish executing. After a few minutes then execute the following command.

    aws cloudformation describe-stacks --stack-name namesblogpoststack --region <region> | grep 'StackStatus'
   

   When the status is reported back as CREATE_COMPLETE then proceed to the next step of importing names.

5. After the cloud formation stack is created successfully then upload the names.csv file to the S3 Bucket used as unique-bucket-name-2 in step 2. This step populates the NamesTable with 50 sample names and will trigger the bigram tables to be populated as well. If you would like to use more names there are two other files, one with 500 names (names500.csv) and one with 1000 names (names1000.csv). The 500 and 1000 name files can take a few minutes to finish importing. You can monitor progress in the DynamoDB graphs for each table and when the FirstNameBiGramIndex table and the LastNameBiGramIndex table stops showing write activity then the import has completed. The DynamoDB tables in the CloudFormation script are configured for on-demand scaling and pricing can be affected by the import process.

   aws s3 cp names.csv s3://<replace with unique-bucket-name-2>
   

Bot Usage

To interact with the Bot navigate to the Lex V2 Console. In the list of Bots, you should now see "NameMatchToDirectoryBot". Click on the Bot which will open the Bot Details page. On the left hand side, under English (US), click on Intents. At the bottom right side of the screen there is a "Test" button. Click on the test button which will open a window named "Test Draft Version". You can now use the textbox or the mic button to give input to the bot. The Bot's workflow will begin if you type "Ready" and hit Enter.