Multi-turn cross-domain natural language understanding systems, building platforms, and methods

A system with a platform for building multi-turn cross-domain natural language understanding (NLU) system for an application, the system comprising:

• at least one processor;
• and
• a memory for storing and encoding computer executable instructions that, when executed by the at least one processor is operative to:
• retrieve multi-domain multi-turn tasks;
• train a first model utilizing the multi-domain multi-turn tasks to predict schematic labeling of a user input to form a schema prediction model;
• retrieve multi-domain multi-turn NLU data;
• and
• train a second model utilizing the multi-domain multi-turn NLU data to form a natural language understanding (NLU) model,
• wherein the schema prediction model and the NLU model form the multi-turn cross-domain NLU system.

A method for building a multi-turn cross-domain natural language understanding (NLU) system, the method comprising:

• training a first model to predict schematic labeling of a user input to form a schema prediction model;
• training a second model to form a NLU model;
• wherein the schema prediction model and the NLU model form the multi-turn cross-domain NLU system that is capable of:
• receiving the user input;
• predicting a schema utilizing the schema prediction model based on the user input;
• constraining the NLU model based on the schema to form a constrained NLU model;
• and
• decoding the user input utilizing the constrained NLU model to determine a response to the user input.