Text analysis concepts
Text analysis is an important aspect of conversational channels that enables a Pega Platform application to intelligently and seamlessly interact with a user in a natural conversational manner. Text predictions and text analyzers examine user input by using natural language processing (NLP), predictive and adaptive analytics, and artificial intelligence to find the best matching response.
Text predictions vs text analyzers
To use text analysis in conversational channels, configure a text prediction or one or more text analyzers for an Pega Intelligent Virtual Assistant™ (IVA) or Pega Email Bot™ in your application.
Text predictions are a new type of prediction in Prediction Studio that provides efficient configuration and monitoring of text analytics for your channels. The system automatically creates a text prediction for every channel that you add in your application. Every text prediction is associated with a text analyzer as its basis. You can open the text analyzer from the Behavior tab in your channel or from the Actions menu in the text prediction.
Text predictions are targeted to replace text analyzers that were used in previous versions of Pega Platform for analyzing messages in conversational channels. For more information about text predictions, see the following articles:
- Text predictions simplify the configuration of text analytics for conversational channels
- Predict customer needs and behaviors through text analytics in your conversational channels
- Analyzing messages with text predictions
Text analyzers are still available in Pega Platform and you can continue to use them. However, to get the best results out of your text analysis and to support more efficient updates, use text predictions.
Text analysis features
Text predictions and text analyzers detect the following information categories:
- The general subject, an intent of an email, text message, or a voice command. An IVA or an email bot links all suggested cases and suggested responses to topics. For example, an email bot can determine that the topic of an email relates to a car insurance, and then open a car insurance business case.
- The text contains proper nouns that fall into a common category, for example, a person, location, date, organization, or ZIP code. You can configure the IVA or the email bot to automatically assign the entities that they detect to properties of a new business case.
- The opinion that a user expresses in an email, a chat text message, or a voice command: positive, neutral, or negative. An email bot can detect a negative sentiment of a user email, and then escalate the issue by automatically forwarding that information to a customer service representative.
- The language of an email, a chat text message, or a voice command. An email bot can detect the language of a user email, perform text analysis in that language by using NLP, and then automatically send a reply in this language to the user.
Text analysis in an email bot
Each text prediction or text analyzer configured for an email bot supports advanced text analysis of email header, body, subject, and attachments, including image files. To perform text analysis of image-based file attachments, you use the Pega optical character recognition (OCR) component that you install from Pega Marketplace on premises.
With text analysis and intelligent email routing, an email bot interprets an email and determines how to correctly respond to a user. This functionality also improves the triage process by creating a correct business case with the help of the email topic.
To refine the text analysis capability for your email bot, you can define multiple models in a text prediction to serve the same or different purposes.
Text analysis in an IVA
A text prediction or text analyzer that you configure for an IVA provides advanced text analysis of user input, including text voice commands. The text prediction helps the system determine the best matching response by using NLP, adaptive analytics, and artificial intelligence.
Previous topic Conversational channel user roles Next topic Exploring text predictions and text analyzers