Amazon Technologies, Inc.

Aaron A Eakin

Ariya Rastrow

Sebastian Lehmann

Peter C Fang

Thomas Park

Yusuf Bootwala

Bjorn Hoffmeister

Nikko Strom

Eli Joshua Fidler

Roland Maximilian Rolf Maas

Diamond Bishop

Rich Wasserman

Sanjeev Mishra

James G Droppo

Chenlei Guo

Xing Fan

Ankur Gandhe

Frederick Victor Weber

Venkata Kishore Nandury

Andrew Oberlin

Angela Sun

Pankaj Sitpure

Richard Wasserman

Gurpreet Singh Chadha

Milind Mukesh Rao

Santosh Kumar Patro

Buddha Puneeth Nandanoor

Anand Mohan

Gopinath Chennupati

Prahalad Venkataramanan

A first neural network (NN) model may generate labels for training a second NN model. The second NN model may represent instances of a NN model operating on multiple different devices (e.g., decentralized user and/or edge devices). The system may include using a 'teacher' model to process data received by one or more of the devices to generate a labeled dataset. The system may use the labeled dataset and a “student” model to calculate gradient data for updating the student model. The student model may be the same or similar to NN model instances operating on the devices. The system may validate the updated student model to determine, for example, whether it exhibits improved performance when processing the newly received data and/or historical data. The system may distribute the validated update to the devices.

Continuous learning for machine learning models

A speech interface device is configured to detect an interrupt event and process a voice command without detecting a wakeword. The device includes on-device interrupt architecture configured to detect when device-directed speech is present and send audio data to a remote system for speech processing. This architecture includes an interrupt detector that detects an interrupt event (e.g., device-directed speech) with low latency, enabling the device to quickly lower a volume of output audio and/or perform other actions in response to a potential voice command. In addition, the architecture includes a device directed classifier that processes an entire utterance and corresponding semantic information and detects device-directed speech with high accuracy. Using the device directed classifier, the device may reject the interrupt event and increase a volume of the output audio or may accept the interrupt event, causing the output audio to end and performing speech processing on the audio data.

Device-directed utterance detection

Techniques are described for automatically identifying and inhibiting suspect electronic communications between users, such as for electronic communications that may reflect fraudulent activities by one or more of the users or that are otherwise inappropriate for one or more reasons. In some situations, the communications may relate to transactions in which the users are involved, such as for users that are buyers and sellers of items interacting via one or more electronic marketplaces. The communications may occur via a communication mechanism provided by an electronic marketplace that acts as an intermediary between the sender of a communication and the one or more intended recipients of the communication. The automatic identification of suspect communications that may be inappropriate includes automatically assessing the communications in one or more ways, and identified suspect communications and any associated inappropriate user activities may be inhibited in various ways.

Inhibiting inappropriate communications between users involving transactions

Described herein is a system for prompting a user for clarification when an automatic speech recognition (ASR) system encounters ambiguity with respect to the user's input. The feedback provided by the user is used to retrain machine-learning models and/or to generate new machine-learning models. Based on the type of ambiguity, the system may determine to retrain one or more ASR models that are widely used by the system or to generate/update one or more user-specific models that are used to process inputs from one or more particular users.

Disambiguation in automatic speech processing

Disclosed are various embodiments related to intercepting, modifying, and analyzing messages being exchanged between two entities. In one embodiment, among others, a method comprises intercepting messages being exchanged between a first entity and a second entity. The intercepted messages are modified to substitute sender information identifying the sender with proxy sender information such that the sender is anonymous to the intended recipient. The messages are also analyzed to determine at least one performance metric associated with the messages being exchanged.

Closed loop communication

Disclosed are various embodiments of a system. In one embodiment, among others, a method comprises intercepting a communication between a first party and a second party in a communication forum. The communication includes first party proxy information as an intended recipient information. The method further comprises accessing a communication pair using the intended recipient information. The intended recipient information is associated with second party proxy information. The second party proxy information is associated with second party information. Additionally, the method comprises determining whether the identity of the sender is valid.

Disclosed are various embodiments of a system. In one embodiment, among others, a method comprises intercepting a communication between a first party and a second party in a communication forum, the communication including a first party proxy information as an intended recipient information. The method further comprises accessing a communication pair using the intended recipient information, the intended recipient information being associated with a second party proxy information, the second party proxy information being associated with a second party information. Additionally, the method comprises determining whether the identity of the sender is valid.

Inhibiting inappropriate communications between users involving tranactions

Growing community of inventors

Redmond, WA, United States of America

Aaron A Eakin