Text predictions simplify the configuration of text analytics for conversational channels
Valid from Pega Version 8.6
Enable text analytics for your conversational channels, such as email and chatbot, by configuring text predictions that manage the text models for your channels. This new type of prediction in Prediction Studio consolidates the AI for analyzing the messages in your conversational channels in one place and replaces the text analyzer rule in Dev Studio.
Through text predictions, you can efficiently configure the outcomes that you want to predict by analyzing the text in your channels:
- Topics (ticket booking, subscription cancellation, support request)
- Sentiments (positive, neutral, negative)
- Entities (people, organizations, airport codes)
- Languages
You can train and build the models that predict these outcomes through an intuitive process, and then monitor the outcomes through user-friendly charts.
For more information, see Predict customer needs and behaviors by using text predictions in your conversational channels.
Upgrade impact
Channels that you configured with text analyzers in the previous version of your system continue to work in the same manner after the upgrade to the current version. When you edit and save the configuration of an existing channel, the text analyzer rule is automatically upgraded to a text prediction. The associated text prediction is now an object where you can manage and monitor the text analytics for your channel. When you create a new channel in the upgraded system, the system automatically creates a text prediction for that channel.
What steps are required to update the application to be compatible with this change?
- Enable the asynchronous model building and reporting in text predictions through job schedulers that use the System Runtime Context (SRC) by adding your application to the SRC.
For more information, see Automating the runtime context management of background processes. - Enable model building in text predictions by configuring background processing nodes.
For more information, see Assigning decision management node types to Pega Platform nodes.
Upgrading to Hazelcast 4.x requires downtime during upgrades to Pega Infinity 8.6
Valid from Pega Version 8.6
Upgrade impact statement
On-premises upgrades of Pega Infinity release 8.4.2 and later to version 8.5.1 or later on Tomcat and PostgreSQL are completed with near-zero downtime. However, upgrading to Hazelcast 4.x requires that you shut down and restart your application server.
What is required to update the application to be compatible with this change?
Hazelcast 3.x is enabled by default. For near-zero downtime upgrades, you do not need to perform any action.
For instructions about upgrading to Hazelcast 4.x, see one of the following topics:
- For near-zero downtime upgrades from Pega Infinity release 8.4.2 or later on Tomcat and PostgreSQL, see "Optional: upgrading to Pega Platform version 8.6: Upgrading to Hazelcast 4.x" in Near-zero downtime Upgrade Guide for Pega Platform version 8.4.2 and later for Tomcat and PostgreSQL.
- For all other upgrades, see "Optional: upgrading to Hazelcast 4.x" in the appropriate upgrade guide.
External data flow rules are removed
Valid from Pega Version 8.6
In previous versions of Pega Platform™, you could configure data flows to run in an external Hadoop environment. The external data flows functionality was deprecated and hidden from view in Pega Platform 8.5. The functionality has been now removed and is no longer available in Pega Platform 8.6.
For more information, see External data flow rules are deprecated.
Improved indexing of StringList and StringGroup property types
Valid from Pega Version 8.6
Search and Reporting Service in Pega Platform™ 8.5 may improperly index StringList and StringGroup property types. As a result, the data model does not include the affected properties.
Upgrade impact
After upgrading to Pega Platform version 8.6, the system requires that the classes with the StringList or StringGroup type are reindexed.
What steps are required to update the application to be compatible with this change?
On the Search Landing Page, manually reindex all the classes that include properties with the StringList or StringGroup types to ensure that all your data is present in the data model. Alternatively, if finding specific instances of classes is difficult, you can reindex all classes in your application.
For more information, see Indexing class data.