Amazon Technologies, Inc.

Shirin Saleem

Kurt W Piersol

Rahul Gupta

Spyridon Matsoukas

Aimee Piercy

Marcello Typrin

Shamitha Somashekar

Chengwei Su

Sankaranarayanan Ananthakrishnan

Kelly James Vanee

Kavya Ravikumar

Timothy Thomas Gray

Kynan Dylan Antos

Jeremy Richard Hill

Jenny Toi Wah Lam

Melanie Chie Bomke Gens

Jonhenry A Righter

Chungnam Chan

Mallory McManamon

Michal Grzegorz Kurpanik

Joe Virgil Fernandez

John Will Crimmins

Luca Soldaini

John Pelak

Sarveshwar Nigam

Yugang Li

Techniques for processing a user input are described. Text data representing a user input is processed with respect to at least one finite state transducer (FST) to generate at least one FST hypothesis. Context information may be required to traverse one or more paths of the at least one FST. The text data is also processed using at least one statistical model (e.g., perform intent classification, named entity recognition, and/or domain classification processing) to generate at least one statistical model hypothesis. The at least one FST hypothesis and the at least one statistical model hypothesis are input to a reranker that determines a most likely interpretation of the user input.

Natural language understanding processing

A natural language understanding system that can determine an overall score for a natural language hypothesis using hypothesis-specific component scores from different aspects of NLU processing as well as context data describing the context surrounding the utterance corresponding to the natural language hypotheses. The individual component scores may be input into a feature vector at a location corresponding to a type of a device captured by the utterance. Other locations in the feature vector corresponding to other device types may be populated with zero values. The feature vector may also be populated with other values represent other context data. The feature vector may then be multiplied by a weight vector comprising trained weights corresponding to the feature vector positions to determine a new overall score for each hypothesis, where the overall score incorporates the impact of the context data. Natural language hypotheses can be ranked using their respective new overall scores.

Contextual natural language processing

A dock device connects participating devices such as a tablet device and an audio activated device, allowing them to operate as a single device. These participating devices may be associated with different accounts, each account being associated with particular 'speechlets' or data processing functions. A natural language understanding (NLU) system uses NLU models to process text obtained from an automatic speech recognition (ASR) system to determine a set of possible intents. A second set of possible intents may then be generated that is limited to those possible intents that correspond to the speechlets associated with the docked device. The intents within the second set of possible intents are ranked, and the highest ranked intent may be deemed to be the intent of the user. Command data corresponding to the highest ranked intent may be generated and used to perform the action associated with that intent.

Natural language understanding with affiliated devices

A speech system may be configured to operate in conjunction with a stationary base device and a handheld remote device to receive voice commands from a user. A user may direct speech either to the base device or to the handheld device. In order to direct speech to the base device, the user first speaks a keyword. In order to direct speech to the handheld device, the user presses a talk control on the handheld device. A dialog may be conducted with the user in multiple turns, where each turn comprises user speech and a speech response by the speech system. The user speech in any given dialog turn may be provided from the base device and/or the handheld device.

Multiple-source speech dialog input

A speech system may be configured to operate in conjunction with a stationary base device and a handheld remote device to receive voice commands from a user. Voice commands may be directed either to the base device or to the handheld device. When performing automatic speech recognition (ASR), natural language understanding (NLU), dialog management, text-to-speech (TTS) conversion, and other speech-related tasks, the system may utilize various models, including ASR models, NLU models, dialog models, and TTS models. Different models may be used depending on whether the user has chosen to speak into the base device or the handheld audio device. The different models may be designed to accommodate the different characteristics of audio and speech that are present in audio provided by the two different components and the different characteristics of the environmental situation of the user.

Source-specific speech interactions

Growing community of inventors

Belmont, MA, United States of America

Shirin Saleem