Skip to main content


         This documentation site is for previous versions. Visit our new documentation site for current releases.      
 

Analyzing messages with text predictions

Updated on July 5, 2022

Text predictions use natural language processing (NLP), predictive and adaptive analytics, and artificial intelligence to analyze incoming messages in your conversational channels, such as email or chat. Text analytics can help you route work, populate properties in business cases, and respond to users with relevant messages.

The system creates a text prediction automatically for each new channel that you create in App Studio. A link to the text prediction is available on the Behavior tab in the channel configuration.

Tip: You can also create a prediction for text analytics in Prediction Studio and add the text prediction to a conversational channel later. For more information, see Creating text predictions.
Text prediction outcomes
The Outcomes tab in a text prediction

You can configure and train the models in a text prediction to predict different aspects of emails, chat messages, or voice commands, and then use that information to automate certain tasks in your application:

Topic
The general subject or intent of a message, such as a request for service or support. For example, a Pega Intelligent Virtual Assistant (IVA) or an email bot can determine that the topic of an email is a request to cancel a flight ticket, and then open a flight cancellation business case.
Entities
Keywords and phrases in a message that the system can assign to specific categories, such as people, locations, dates, organizations, and postal codes. You can configure an IVA or an email bot to automatically assign the detected entities to properties in a business case.
Sentiment
The attitude or opinion that a user expresses in a message: positive, neutral, or negative. An email bot can detect negative sentiments in an email, and then escalate the issue by automatically forwarding that information to a customer service representative.
Language
The language of a message. An email bot can detect the language of an email, and then automatically respond to the user in that language.

The following diagram shows the high-level workflow for configuring a text prediction:

Text prediction workflow
A configuration workflow for a text prediction
Before you begin: In a non-production environment, create an email or IVA channel in App Studio. For more information, see Creating an Email channel and Creating a Digital Messaging channel.

    Configuring outcomes in a non-production environment

  1. Add topics that you want to detect in incoming messages in your conversational channels:
  2. Add entities that you want to detect in incoming messages in your conversational channels.
    For more information, see Adding entities to text predictions.
  3. Add, review, and approve the training data for topics, sentiments, and entities.
  4. Optional: Add models to the text prediction and configure their settings.
    For more information, see Managing models in text predictions.
  5. Optional: Modify the general settings of your text prediction and manage its models.
  6. Building models and testing predictions

  7. Rebuild the models in the text prediction.
    For more information, see Building models in text predictions.
  8. Test the text prediction.
    For more information, see Testing text predictions.
  9. Deploying to Production

  10. Deploy the text prediction to the Production environment.
    For more information, see Moving applications between systems.
  11. Monitoring text predictions

  12. In the production environment, monitor the performance of your text prediction to identify opportunities to improve the configuration of text analytics for your conversational channels.
    For more information, see Monitoring text predictions.
What to do next: Automate tasks by configuring your application behavior based on the results of the text analysis. For example, you can route emails to specific work queues based on the detected topics as shown in the following video:

To learn more about text analytics and email routing, see the following Pega Academy module: Text analytics for email routing.

To get hands-on experience in configuring email routing by using text analytics, take up the following Pega Academy challenge: Training a topic model to improve email routing.

Have a question? Get answers now.

Visit the Support Center to ask questions, engage in discussions, share ideas, and help others.

Did you find this content helpful?

Want to help us improve this content?

We'd prefer it if you saw us at our best.

Pega.com is not optimized for Internet Explorer. For the optimal experience, please use:

Close Deprecation Notice
Contact us