To have greater control of the quality of machine learning in your Pega Intelligent Virtual Assistant™ (IVA) and Pega Email Bot™, you can now copy training data and the unbuilt natural language processing (NLP) model from one IVA or Email Bot environment to another. The ability to copy data between IVA and Email Bot environments results in a higher number of triaged emails and greater accuracy of entity detection, and makes it easier to maintain IVA and build its artificial intelligence.
Entities are detected by the system in a chat conversation, including attachments, to help respond to the user correctly, and consist of proper nouns that fall into a commonly understood category such as a person, organization, location, date, and for example, can also be a Social Security number, an email address, or a ZIP code.
The following figure shows the IVA Web Chatbot Training data tab.
You copy data by exporting the data from an IVA or Email Bot environment to a RAP file, and then importing the RAP file from your local disk to another IVA environment. For example, you can generate training data in a production environment for your IVA or Email Bot, and then copy it to a development or a QA environment. You can either export or import the reviewed training data or all the training data, including training records that are not reviewed. Data that is exported includes entities that were added to the text analyzer but were not built.
For more information, see Copying training data to another environment in IVA for Facebook, Copying training data to another environment in IVA for Web Chatbot, and Copying training data to another environment in Email Bot.