Abstract: A multi-channel transformer acoustic model that processes a plurality of audio signals output by microphones of a microphone array and outputs probabilities for acoustic units of an utterance represented in the audio signals. The audio signals represent the individual microphones' respective capturing of the utterance. The multi-channel model may perform self-attention on embeddings of the audio signals and then cross-channel attention across the attended audio signals. The cross-channel attention may involve processing of signals relative to each other to model the relationships across channels within and across time frames. The multi-channel model may include a transducer to perform processing frame-by-frame.

Abstract: A system configured to improve track selection while performing audio type detection using sound source localization (SSL) data is provided. A device processes audio data representing sounds from multiple sound sources to determine SSL data that distinguishes between each of the sound sources. The system detects an acoustic event and performs SSL track selection to select the sound source that corresponds to the acoustic event based on input features. To improve SSL track selection, the system detects current conditions of the environment and determines adaptive weight values that vary based on the current conditions, such as a noise level of the environment, whether playback is detected, whether the device is located near one or more walls, etc. By adjusting the adaptive weight values, the system improves an accuracy of the SSL track selection by prioritizing the input features that are most predictive during the current conditions.

Abstract: Systems and methods are provided for prioritized restoration of multiple network connections that have been interrupted. Connection state data (e.g., quantity of pending IO requests) may be used to prioritize reestablishment of some connections over others. In addition, the connections may be reestablished using “in-band” handshakes over the previously-established network connections.

Abstract: A text-to-speech (TTS) system may be configured to imitate characteristics of a target voice based on a limited dataset. The TTS system may include a machine learning model pre-trained using a synthetic parallel dataset and fine-tuned using examples of the target voice. A TTS component trained using a large single-speaker dataset may be used to generate the synthetic parallel dataset based on a multi-speaker dataset. The synthetic parallel dataset may include target audio data representing speech in the multi-speaker dataset and predicted audio data generated by the TTS component based on transcripts of the speech. The machine learning model may be pre-trained using the synthetic parallel dataset and fine-tuned using audio data representing target voice speech and predicted audio generated by the TTS component based on transcripts of the target voice speech. The trained model may be used to modify synthetic speech to approximate the characteristics of the target speech.

Abstract: Systems and techniques for producing image-based radiology reports including contextual cropping of image data and radiologist supplied notes and annotations are provided herein. Computer vision and natural language processing algorithms may enable processing of image data and language inputs to identify objects associated with annotations, aid in cropping the image data according to the annotations and object identification and in producing a final text and image laden report.

Abstract: A scalable encoding architecture is able to support multiple encoder versions without requiring new versions of the architecture to be generated and supported for each encoder version. An encoding profile template can be generated that includes the available settings for an encoder. A versioned schema can be generated that defines the settings for a particular encoder version. The schema can be persisted such that when a new channel or media stream is to be generated, the schema can be loaded and its values validated in order to cause the encoder to encode the media content using the appropriate values. An interface presented to the user to generate the template or channel can be dynamically generated such that only the applicable and/or relevant settings for a particular encoder version are displayed.

Abstract: In accordance with input received via a programmatic interface, a level of isolation at which a software container is to be executed is determined. Based on the level of isolation, a category of virtual machines of a virtualized computing service is selected for executing the software containers. The selected category differs from other categories in at least the number of devices emulated for virtual machines of the categories. The software container is run within a virtual machine of the selected category.

Abstract: Quorum-based access control management may be implemented. Quorum controls may be created for determining whether to perform or deny access control operations to perform privileged tasks. When an access control operation is received, approval of the operation may be requested from members for the quorum control. If a policy for the quorum control is satisfied by approval responses, then approval to perform the access control operation may be provided.

Abstract: Aspects described herein relate to securely performing cross-boundary backup operations. A service of a computing resource service provider may enable backup operations between a source account and a destination account of an organization based at least in part on a security policy allowing such operations.

Abstract: Optimistic concurrency may be implemented for multiple writers to a database. Writes to a database from different writers may succeed unless a conflict is detected at the data store for the database. For detected conflicts, a request to resolve the conflict between writes may be submitted to a primary writer, which may determine and provide a resolution of the conflict to the conflicting writers. The data store may then be updated to commit the selected write identified by the resolution of the conflict.

Abstract: Systems and methods are provided for exposing disparate storage systems and technologies to compute instances through a unified file system interface that is agnostic to the underlying storage systems and technologies. A file system namespace service may provide the ability to compose a file system from different storage technologies and systems (e.g., block storage, object storage, etc.). The file system namespace service may also provide the ability to layer different storage services (e.g., by treating local block storage volume as a write back cache, by treating a file system as a read-through cache in front of an object storage service, etc.), and/or the ability to bind virtual files to other compute instances and/or storage systems.

Abstract: A commonly designed processor heat exchanger decouples the mounting hardware used to mount the heat exchanger to a processor from the heat exchanger itself. This allows a single heat exchanger design to be mounted to various different types of processors using processor customized mounting brackets that engage with flanges extending out from a body of the heat exchanger.

Abstract: As agents move about a materials handling facility, tracklets representative of the position of each agent are maintained along with a confidence score indicating a confidence that the position of the agent is known. If the confidence score falls below a threshold level, image data of the agent associated with the low confidence score is obtained and processed to generate one or more embedding vectors representative of the agent at a current position. Those embedding vectors are then compared with embedding vectors of other candidate agents to determine a set of embedding vectors having a highest similarity. The candidate agent represented by the set of embedding vectors having the highest similarity score is determined to be the agent and the position of that candidate agent is updated to the current position, thereby re-identifying the agent.

Abstract: Techniques for presenting an alternative input representation to a user for testing and collecting processing data are described. A system may determine that a received spoken input triggers an alternative input representation for presenting. The system may output data corresponding to the alternative input representation in response to the received spoken input, and the system may receive user feedback from the user. The system may store the user feedback and processing data corresponding to processing of the alternative input representation, which may be later used to update an alternative input component configured to determine alternative input representations for spoken inputs.

Abstract: Techniques for storing and using multi-session context are described. A system may store context data corresponding to a first interaction, where the context data may include action data, entity data and a profile identifier for a user. Later the stored context data may be retrieved during a second interaction corresponding to the entity of the second interaction. The second interaction may take place at a system different than the first interaction. The system may generate a response during the second interaction using the stored context data of the prior interaction.

Abstract: Systems, methods, and computer-readable media are disclosed for dynamic voice search transitioning. Example methods may include receiving, by a computer system in communication with a display, a first incoming voice data indication, initiating a first user interface theme for presentation at a display, wherein the first user interface theme is a default user interface theme, and receiving first voice data. Example methods may include sending the first voice data to a remote server for processing, receiving an indication from the remote server to initiate a second user interface theme, and initiating the second user interface theme for presentation at the display.

Abstract: Techniques are described for providing logical networking functionality for managed computer networks, such as for virtual computer networks provided on behalf of users or other entities. In some situations, a user may configure or otherwise specify a network topology for a virtual computer network, such as a logical network topology that separates multiple computing nodes of the virtual computer network into multiple logical sub-networks and/or that specifies one or more logical networking devices for the virtual computer network. After a network topology is specified for a virtual computer network, logical networking functionality corresponding to the network topology may be provided in various manners, such as without physically implementing the network topology for the virtual computer network. In some situations, the computing nodes may include virtual machine nodes hosted on one or more physical computing machines or systems, such as by or on behalf of one or more users.

Abstract: A system that is capable of resolving anaphora using timing data received by a local device. A local device outputs audio representing a list of entries. The audio may represent synthesized speech of the list of entries. A user can interrupt the device to select an entry in the list, such as by saying “that one.” The local device can determine an offset time representing the time between when audio playback began and when the user interrupted. The local device sends the offset time and audio data representing the utterance to a speech processing system which can then use the offset time and stored data to identify which entry on the list was most recently output by the local device when the user interrupted. The system can then resolve anaphora to match that entry and can perform additional processing based on the referred to item.

Abstract: A simulation environment (e.g., multi-player game) may be hosted by multiple nodes of a provider network. The number of nodes may increase or decrease, depending on the load on the simulation. Each node may be assigned a root key that includes 3D coordinates. The nodes may maintain a circular keyspace that includes the x-axis coordinates, the y-axis coordinates, and the z-axis coordinates, according to a consistent hashing algorithm. A given node may have any number of successor or predecessor nodes along each axis keyspace. If load on a particular node increases above a threshold amount, a new node may be allocated and insert itself into the 3D keyspace, taking over authority for any entities that are located within a portion of the volume that was previously managed by the particular node.

Abstract: A content streaming system and methodology for facilitating the management of content streaming. A content delivery service receives at least an original content stream and multiple input contents from an original content provider. Additionally, the content delivery service processes the received input content streams, buffers the original and supplemental input content streams, and switches between the available input content streams for seamless and fast broadcasting. The decision to switch between a second content and a third is determined by an indication that a third input content is available in the buffer.