Package org.nlpcraft.model.intent

Contains intent-based input matcher.

See: Description

Interface Summary

Interface	Description
NCIntentSolver.COMBINATOR	Marker type for built-in combinators.
NCIntentSolverContext	A context that can be passed passed into the callback of the matched intent.

Class Summary

Class	Description
NCIntentSolver	Main entry point for intent-based user input matcher.
NCIntentSolver.AND	AND-combinator.
NCIntentSolver.CONV_INTENT	Convenient adapter for conversational unordered intent.
NCIntentSolver.INTENT	An intent is a collection of terms defining a formal template for user input.
NCIntentSolver.NAND	NAND-combinator.
NCIntentSolver.NON_CONV_INTENT	Convenient adapter for non-conversational unordered intent.
NCIntentSolver.NOR	NOR-combinator.
NCIntentSolver.NOT	NOT-combinator.
NCIntentSolver.OR	OR-combinator.
NCIntentSolver.RULE	Declarative predicate over NCToken.
NCIntentSolver.TERM	Term is a building block of the intent.
NCIntentSolver.XNOR	XNOR (exclusive NOR)-combinator.
NCIntentSolver.XOR	XOR (exclusive OR)-combinator.

Exception Summary

Exception	Description
NCIntentSkip	Control flow exception to skip current intent.

Annotation Types Summary

Annotation Type	Description
NCIntent	Annotation to define an intent on the method serving as its callback.
NCIntentRef	Annotations referencing an intent defined externally.
NCIntentTerm	Annotation to mark callback parameter to receive term's tokens.

Package org.nlpcraft.model.intent Description

Contains intent-based input matcher.

Intent-based input matcher (or solver) is intended for simpler and more declarative implementation of model's query(NCQueryContext) method. This is the main method in the data model and it is responsible for converting parsed user input into an action. It cannot be defined in the external JSON or YAML model definition and, therefore, has to be defined in the code. Although this method can be implemented manually by encoding the logic of parsing query context and its list of parsing variants - it is often more complex and error-prone than necessary.

Intent solver is based on an idea of adding one or more user input templates, also known as intents, and letting the solver select the best matching intent for a given the user input. Each intent defines a pattern of the user input and associated action to take when that pattern is selected. While selecting the best matching intent the solver uses sophisticated NLP analysis and conversation context for anaphoras resolution. Among multiple detected intents the solver will always select the most specific one.