Terms & concepts

Get acquainted with common terms related to conversational AI and your use of NLX

Common workspace terms

Resource	Definition
NLP	Natural Language Processing (NLP); processes a user's lexical queries for matching to training data and slot values
NLU	Natural Language Understanding (NLU); built-into the NLX product, the NLU keeps track of conversation state and interfaces with backend and frontend services
LLM	Large Language Model; comprehends and generates human language responses from large deep learning models
Applications	The conversational AI application that users interact with that is defined in your workspace by the channels and intents you assign it
Channels	The conversation method (voice, chat) and communication system (Amazon Connect, Twilio, etc.) your application uses to converse with users
Intents	The automated task that follows a business action to handle a specific user query (e.g., book a flight, check showtimes)
Intent flows	A workflow constructed of multiple nodes that defines the process for completing a business action with logic and messaging
Training phrases	The training data provided to AI models to understand user intent and invoke the appropriate intent workflow
Slots	Slots are parameters extracted from users that are required for completing a process outlined in an intent workflow. They may be matched to fixed values (custom), or open-ended (using the NLX system slots)
Data requests	APIs that return a response when invoked in an intent flow (e.g., hotel room availability, user name, movie showtimes)
Actions	APIs that trigger an outbound event and simply return a success or failure status
Lifecycle hooks	APIs that represent extendable points in the conversation: Conversation start Escalation Conversation end Data request streaming (streams State modifications applied to Data requests when invoked in conversation)
Voice+ scripts	Multimodal conversation experiences that synchronize voice with externally-hosted digital assets
Knowledge bases	Resource for creating a digital repository of knowledge that can answer common queries from users without the need to construct independent intents to address
History	Displays conversation sessions with deployed bots that also shows NLP information, including NLP confidence score and intents matched
Training	Displays a list of all user utterances that had low NLP confidence scores or triggered the Unknown behavior
Utterances	User inputs that may be spoken or written
Analytics	Custom or templated dashboards sharing key metrics for application performance, API performance, user engagement, and more
Context variables	Method for sharing information between intent flows in a single conversation session that can be statically set during a conversation's start or dynamically updated during the conversation
Secrets	Sensitive or secret values, such as URL strings, API keys, passwords, etc