Bot Framework middleware component for the Amazon AWS Comprehend service, including sentiment analysis, language detection, entity extraction, and keyword recognition.
npm install @botbuildercommunity/middleware-aws-comprehendThe AWS Comprehend Middleware offers Bot Framework middleware components for the AWS Comprehend API. You will need an AWS account, as well as a Comprehend resource created on AWS. Take a look at the Amazon AWS Comprehend documentation for details.
npm install @botbuildercommunity/middleware-aws-comprehend --save
This package assumes that your AWS credentials are set in the environment variables. The following environment variables are required:
AWS_ACCESS_KEY_ID
AWS_SECRET_ACCESS_KEY
All middleware is created and used in the same way. For example, for sentiment analysis, import the SentimentAnalysis class from the package, and add it to your bot adapter:
``typescript
import { SentimentAnalysis } from '@botbuildercommunity/middleware-aws-comprehend';
adapter.use(new SentimentAnalysis());
`
When used, the turnState on the TurnContext will have a property named sentimentScore between 0 and 1. A full example can be seen in the sample bot test file.
> Differences from Cognitive Services and Watson: AWS Comprehend does not return a 0 - 1 response on the sentiment. Instead, it returns 0 - 1 for each possible sentiment. You can access the sentiment through the sentimentType property on the turnState.
By default, the middleware assumes an East US region and English as the language. You can modify this by passing in AWSComprehendOptions to the SentimentAnalysis:
`typescript
import { SentimentAnalysis } from '@botbuildercommunity/middleware-aws-comprehend';
adapter.use(new SentimentAnalysis({ apiVersion: '2017-11-27', region: 'us-east-1', lang: 'en' }));
`
Supported middleware classes include:
* SentimentAnalysisLanguageDetection
* KeyPhrases
* EntityExtraction
*
In each case, properties are added to the turnState of the TurnContext. You can retrieve them in your bot via:
* context.turnState.get('sentimentScore') //This is a number for SentimentAnalysiscontext.turnState.get('sentimentType')
* //This is a string for SentimentAnalysiscontext.turnState.get('language')
* //This is a string for LanguageDetectioncontext.turnState.get('keyPhrases')
* //This is an array of strings for KeyPhrasescontext.turnState.get('textEntities')
* //This is an array of strings for EntityExtraction`