A multi-language chatbot with automated translation

Description

This is a sample chatbot application using Amazon Lex and Amazon Translate to automatically translate incoming requests, process them and translate the responses back to their original language.

It is based on claudia-bot-builder and it is compatible with many messaging platforms, such as Facebook Messenger, Telegram, Skype, Slack and also Amazon Echo (Alexa). claudia-bot-builder runs on Amazon Lambda and Amazon API Gateway.

Read this article to find out more.

Caveats

• This chatbot is meant to prototype a new service or product and run the first iterations over it. Although it can scale virtually infinitely using the Amazon platform, for the long run it is most probably better to switch to human-curated translations and save on the requests issued to Lex and Translate, that are billed per-use.

• The AWS free tier SHOULD allow you to test this bot for free, but please read its terms and conditions carefully. You may be charged for AWS services usage. In particular, Lex and Translate free tiers only last for 12 months after the first request (at the time of writing). So if your account is older than that you may have already used up your free tier. Also, all services subject to the free tier do enforce usage limits for it. Make sure that your usage is within those limits. Be safe and set an AWS Budget to detect unwanted or excessive charges early on. You are responsible for that.

• The training data (city names) provided to the chatbot for the Location slot is very minimal and not enough for the machine learning to properly determine its nature, so stick to the provided list of cities when testing the bot. Other cities might not work.

Requirements

• Node.JS v10.x.x

Deployment

Claudia.JS notes

This bot is built on claudia-bot-builder and uses Claudia.JS for deployments on AWS. Since version 2 of the bot framework was released, Claudia.JS was decoupled from the framework and MUST be installed separately. Claudia.JS can be either installed as a global package or a local dependency. In order to make the distribution of this package easier for users and for it to coexist with other versions of Claudia.JS, it has been included as a local dependency.

First deployment

1. Creating the Lex bot

Zip the aws/lex/WeatherBot_Export.json file and import it using the Lex console:

1. Sign in to the AWS Management Console and open the Amazon Lex console, Bots page.
2. From Actions, choose Import.
3. From Import Bot, choose the .zip archive that contains the JSON file that contains the bot to import.
4. Press Import.
5. Press Build to build the bot.
6. Press Publish and name the deployment: Development.

2. Deploying the backend

Install the required dependencies first:

$ npm install

Run the following command, replacing {AWS_REGION} with your AWS region of choice:

$ ./node_modules/.bin/claudia create \
    --region {AWS_REGION} \
    --api-module bot \
    --version development \
    --policies aws/iam/policies.json \
    --set-env-from-json aws/lambda/env.json

If you have multiple AWS profiles configured on your machine, specify the profile

$ AWS_PROFILE={PROFILE_NAME} \
  ./node_modules/.bin/claudia create \
    --region {AWS_REGION} \
    --api-module bot \
    --version development \
    --policies aws/iam/policies.json \
    --set-env-from-json aws/lambda/env.json

3. Connecting the bot to a messaging platform

For detailed information on connecting a bot to a messaing platform, please see the claudia-bot-builder Hello World Chat Bot doc.

Telegram

To get started quickly and test a bot, Telegram is a very good place to start from. Creating and configuring a bot takes only a few steps.

1. Create a bot on Telegram
2. Deploy the chatbot as described in First deployment above
3. Run

   $ ./node_modules/.bin/claudia --configure-telegram-bot
   

   and follow the instructions shown on the screen.

Usage

This bot uses the MetaWeather API to deliver the weather forecast. You can ask it:

How is the weather in Milan?
How will the weather be like in Boston on monday?

in any language supported by Amazon Translate that can be translated to English. Please read the Caveats paragraph above and see the aws/lex/WeatherBot_Export.json file for the list of cities that are safe to use in this demo.

Tests

Run:

$ npm install
$ npm test

License

This code is released under the MIT license. Please read it carefully before using it.