An Engine is an abstraction over various Natural Language Processing (NLP) services. It contains provider-specific configuration about connecting to the NLP service and interpretation of the service responses.
Engines can be managed in the Operator platform UI under Advanced Admin's OmniGuide tab.
Engines are required for using all OmniGuide features:
- Operator Assistants
- Virtual Assistants
Interpretation of NLP responses
Every NLP response falls into one of the following confidence categories:
- fulfilled: The engine understands Visitor's intent and is confident in its response.
- elicit intent: The engine requires more information to understand Visitor's intent.
- unusable: The engine is unable to provide a meaningful response.
In addition to the confidence category, the Engine response can be one of the following commands:
- suggestion: Text that is suitable to send directly to the Visitor
- prompt: Textual notification about how the Operator should proceed with the conversation
- transfer: JSON payload indicating a Transfer to another Operator or to a Queue. For more information, read our step-by-step guide how to transfer engagements.
Operator Assistants and Virtual Assistants base their actions on both the confidence category and the command of the response. Both Assistants only handle a subset of the commands, it's up to the NLP service and the Client Engineers (with help from OmniGuide metrics) to come up with an useful setup.
Currently Supported NLP services
- Microsoft Bot Framework
- IBM Watson
- AWS Lex