lex-dynamodb

Use Lex to discern intent of online ads and chat conversations

Comment

Inspiration

Polaris Project On-Ramps paper. Amazon Lex.

What it does

Logs intent of an add or chat message to DynamoDB if Lex matches to a human trafficking event or any other intent.

How I built it

JavaScript, DynamoDB, Lambda, Lex, and my first use of AWS SAM.

Challenges I ran into

Publishing an app using SAM.

Accomplishments that I'm proud of

Getting an application published through SAM and using Lex to analyze intent of an ad or message.

What I learned

How to publish an application via SAM

What's next for lex-dynamodb

A front end application to more easily review ads or messages that have been flagged.

Built With

  • dynamodb
  • lambda
  • lex
Share this project:

Updates

posted an update —

Testing instructions:

  1. Deploy dynamodb-lex via the SAM repository. Remember the table name you choose.
  2. Create a Lex bot named DetectTrafficker or anything you like.
  3. Create an intent called "LogIntent" with the following utterances as an example: "great opportunity to make great new friends and see all of the country" "Want to get out and see the country while getting paid cash everyday and having fun"
  4. Under Fulfillment, select "AWS Lambda function" and then choose "logIntent" which should have created by the dynamodb-lex SAM build.
  5. It should ask to allow the bot to execute logIntent. Allow it.
  6. Click "Build".
  7. To test connectivity to the lambda function type "ping". If all is working well you will see a "Pong!".
  8. In the "Test bot" console, verify that a non trafficking ad intent like "make money working from home stuffing envelopes" is not recognized.
  9. Put in a phrase that matches a trafficking intent but doesn't necessarily exactly match the utterances. I.e. make money and see the country
  10. The latter should be recognized and you should see a response that says "Logged!"
  11. Open the table in DynamoDB and verify that the input text was logged.

Log in or sign up for Devpost to join the conversation.