A call center has English transcripts containing only the customer's remarks.
It wants to route each request to Billing, Delivery, or Cancellations and track the sentiment of those remarks in a separate field.
Which two tasks produce the required outputs? Choose two.
Use text classification to select the routing category and sentiment analysis to estimate the tone of the customer's remarks. Requests sent to the same team can express different sentiments.
Detailed explanation
A translation renders the conversation in another language. The translated text alone does not assign routing or sentiment labels.
Predict one of the predefined team labels from the request. The application uses that label to route the request.
Estimate whether the customer's remarks express positive, negative, neutral, or mixed sentiment.
A summary shortens the conversation. The summary text alone does not supply both the routing label and the sentiment label.
Check official documentation
Compare the following points with the official documentation.
Compare the outputs of Amazon Comprehend custom classification and DetectSentiment in the AWS docs. Assign a routing label and a sentiment label to a customer thanking the team for correcting an invoice and to a customer complaining about a duplicate charge.
Official references
- Official documentation 1
- Official documentation 2
- Official documentation 3
- Official documentation 4
- Official documentation 5
What to confirm
Both requests illustrate the Billing routing label. Thanks for the correction illustrate positive sentiment, while a complaint about the duplicate charge illustrates negative sentiment.
Key points
- Text classification
- Sentiment analysis
- Routing and sentiment labels
Notes
- Environment: Paper exercise with AWS documentation and fictional customer remarks
- Use the team label for routing and the sentiment label for reporting. Sentiment intensity alone does not identify the right team.
- Check each feature's supported languages before extending the design beyond English. Custom classification and sentiment analysis have different language coverage.
Foundation review
Routing labels
For Billing, Delivery, and Cancellations in Amazon Comprehend, train a custom classifier on examples labeled with those teams.
Sentiment of the remarks
Amazon Comprehend DetectSentiment returns the overall sentiment of the input text and confidence scores for each label. Separate the customer's remarks before analysis if only their sentiment is required.