Amazon Technologies, Inc.

Felix Xiaomeng Wu

Rohan Mutagi

Rongzhou Shen

Pablo Carballude Gonzalez

Neelam Satish Agrawal

Vibhunandan Gavini

Manish Dutt Sharma

Vikram Kumar Gundeti

Frédéric Johan Georges Deramat

Vijitha Raji

Manuel Jesus Leon Rivas

Ye He

Noel Evans

Miguel Alberdi Lorenzo

Lev Danielyan

Nancy Yi Liang

Philip Alexander Lee

Sai Sailesh Kopuri

Shyam Sunder Kumar

Rahul Ravu

John Dunavent

Daniel Paludi

Luying Pan

Philippe Di Cristo

Jiang Xiang

Yining Bao

Wenjun Xiong

Ellen Hong

Chintan Gohil

Luis Pantoja

Techniques for outputting supplemental content are described. A system may receive input data corresponding to a user input, and determine and present output data responsive to the user input. After causing the output data to be presented, the system may determine supplemental content is to be presented. Based on this, the system may determine first presentation data representing first supplemental content is to be visually presented, and second presentation data representing second supplemental content is to be audibly presented. The system may use a machine learning model to determine the first supplemental content is to be presented instead of the second supplemental content. The system may thereafter cause a device to use the first presentation data to visually present the first supplemental content.

Output of visual supplemental content

Techniques for filtering the output of supplemental content are described. When a supplemental output system (e.g., a supplemental content system or notification system) receives supplemental content for output, the supplemental output system sends a user identifier (of the recipient user) and the supplemental content to separately implemented filtering component. The filtering component uses a machine learning (ML) model to determine a topic of the supplemental content. The filtering component determines whether the supplemental content should not be output based on the ML model-determined topic, one or more guardrail policies of the supplemental output system, and user frustration data regarding previously output supplemental content. Use of the ML model to determine the topic prevents a content publisher from surreptitiously associating supplemental content with a specific topic in an effort to bypass topic-based output guardrails.

Proactive supplemental content output

Configurable core domains of a speech processing system are described. A core domain output data format for a given command is originally configured with default content portions. When a user indicates additional content should be output for the command, the speech processing system creates a new output data format for the core domain. The new output data format is user specific and includes both default content portions as well as user preferred content portions.

Configurable output data formats

A system is provided for determining when supplemental content unresponsive to a user input is to be presented during a user interaction. The system determines an action, performance of which, may trigger output of supplemental content. The system may determine when, during a user interaction corresponding to performance of the action, supplemental content may be presented. The system may use constraint data indicating a device type via which the content may be presented, and a time duration during which the supplemental content may be presented. If the constraint data is satisfied, the system may determine to present the supplemental content.

Supplemental content placement for natural language interfaces

Techniques for performing tests by a speech processing system are described. A speech processing system may apply a hash function to a group of user identifiers, a content provider identifier, and time data to segment the group of user identifiers into non-holdout, holdout control, and holdout treatment groups. A non-holdout group represents users not subject to a test for a duration of time. A holdout control group represents users subject to a test for the duration of time, but which do not receive unresponsive content during the duration of time. A holdout treatment group represents users subject to a test and which receive unresponsive content over the duration of time. User activity of the holdout control and holdout treatment groups may be compared to measure an effectiveness of the unresponsive content. The measured effectiveness may be used to determine a ranking of the content provider at runtime.

Content selection by natural language processing systems

Techniques for synchronously outputting content by one or more devices are described. A system may receive a user command and may receive content responsive to the command from an application(s). The content may include various kinds of data (e.g., audio data, image data, video data, etc.). The system may also receive a presentation framework from the application, with the presentation framework indicating how content responsive to the input command should be synchronously output by one or more devices. The system determines one or more devices proximate to the user, determines which of the one or more devices may be used to output content indicated in the presentation framework, and causes the one or more devices to output content in a synchronous manner.

Multi-modality presentation and execution engine

Systems and methods for proactively deleting applications and/or deactivating permissions granted thereto are described. A system determines a user's usage pattern of an application. The system also determines a standard usage pattern associated with the application. The standard usage pattern may be determined from application usage patterns of multiple users of the system. When the user's usage pattern deviates from the standard usage pattern, the system may cause the application to be deleted from one or more of the user's devices. The system may also deactivate user permissions associated with the application when the user's usage pattern deviates from the standard usage pattern. The system also outputs an indication to the user soliciting whether the system to re-download the application and/or reactivate the permissions.

Proactively limiting functionality

Systems, methods, and devices for dynamically authenticating a user are disclosed. A speech-controlled device captures a spoken command, and sends audio data corresponding thereto to a server. The server determines the audio data includes a spoken command to receive content, and therefrom determines a source storing the content. The server also determines threshold user authentication confidence score data associated with the content source. Based at least in part on the threshold user authentication confidence score data, the server determines a user authentication technique, and a device configured to capture user authentication data. The server determines user authentication confidence score data using user authentication data received from the device, and determines weighted user authentication confidence score data therefrom. If the weighted user authentication confidence score data satisfies the threshold user authentication confidence score data, the server receives the requested content from the content source.

Dynamic user authentication

Growing community of inventors

Seattle, WA, United States of America

Felix Xiaomeng Wu