Parsers for deriving user intents

A computing device, comprising:

• a logic processor;
• and
• a storage device holding instructions executable by the logic processor to:
• analyze with a feeder parser a surface form of a first user input;
• derive with the feeder parser a first user intent underlying the surface form;
• use the first user intent to generate a first output via one or more output devices;
• after generating the first output, receive at least one type of additional data related to the first output and selected from additional user input data and context information;
• use the at least one type of additional data related to the first output to determine an estimated user intent confidence value of the first user intent;
• provide the surface form the first user intent, and the estimated user intent confidence value to a machine learning-based parser;
• use at least the surface form the user intent, and the estimated user intent confidence value to enhance a training set of the machine learning-based parser;
• use the estimated user intent confidence value of the first user intent and the at least one type of additional data related to the first output to derive a second user intent;
• and
• use the second user intent to generate a second output via the one or more output devices.

An intelligent assistant system comprising a computing device, the computing device comprising:

• a logic processor;
• and
• a storage device holding instructions executable by the logic processor to:
• receive audio data of a spoken utterance from another computing device;
• convert the audio data into parsable data;
• analyze with a feeder parser a surface form of the spoken utterance;
• derive with the feeder parser a first user intent underlying the surface form;
• cause the other computing device to broadcast a first message or perform a first action via another device that carries out the first user intent;
• after causing the other computing device to broadcast the first message or perform the first action via the another device, receive at least one type of additional data related to the first message or the first action and selected from additional user input data and context information;
• use the at least one type of additional data related to the first message or the first action to determine an estimated user intent confidence value of the first user intent:
• provide the surface form the user intent, and the estimated user intent confidence value to a machine learning-based parser;
• use at least the surface form the user intent, and the estimated user intent confidence value to enhance a training set of the machine learning-based parser;
• use the estimated user intent confidence value of the first user intent and the at least one type of additional data related to the first message or the first action to derive a second user intent;
• and
• use the second user intent to broadcast a second message or perform a second action via the another device that carries out the second user intent.