Abstract: A method for training a speech recognition model includes obtaining sample utterances of synthesized speech in a target domain, obtaining transcribed utterances of non-synthetic speech in the target domain, and pre-training the speech recognition model on the sample utterances of synthesized speech in the target domain to attain an initial state for warm-start training. After pre-training the speech recognition model, the method also includes warm-start training the speech recognition model on the transcribed utterances of non-synthetic speech in the target domain to teach the speech recognition model to learn to recognize real/human speech in the target domain.

Abstract: A method includes receiving a scan request requesting to scan a set of network-connected assets designated for a network scan. For each respective network-connected asset, the method includes scanning, at a network security scanner using a first scanning privilege level, the respective network-connected asset. The method includes determining, based on the scan using the first scanning privilege level, whether the respective network-connected asset has a vulnerability. In response, the method includes scanning, at the network security scanner using a second scanning privilege level, the respective network-connected asset. The second scanning privilege level defines a lower level of access the network security scanner has than the first scanning privilege level. The method includes determining, based on the scans, an exposure level of the vulnerability. The method includes reporting the exposure level of the vulnerability to a user of the respective network-connected asset.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing instance segmentation by detecting and segmenting individual objects in an image. In one aspect, a method comprises: processing an image to generate data identifying a region of the image that depicts a particular object; obtaining data defining a plurality of example object segmentations; generating a respective weight value for each of the example object segmentations; for each of a plurality of pixels in the region of the image, determining a score characterizing a likelihood that the pixel is included in the particular object depicted in the region of the image using: (i) the example object segmentations, and (ii) the weight values for the example object segmentations; and generating a segmentation of the particular object depicted in the region of the image using the scores for the pixels in the region of the image.

Abstract: A method includes receiving acoustic features of a first utterance spoken by a first user that speaks with typical speech and processing the acoustic features of the first utterance using a general speech recognizer to generate a first transcription of the first utterance. The operations also include analyzing the first transcription of the first utterance to identify one or more bias terms in the first transcription and biasing the alternative speech recognizer on the one or more bias terms identified in the first transcription. The operations also include receiving acoustic features of a second utterance spoken by a second user that speaks with atypical speech and processing, using the alternative speech recognizer biased on the one or more terms identified in the first transcription, the acoustic features of the second utterance to generate a second transcription of the second utterance.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for migrating data from a first storage system to a second storage system. That data is for a process (e.g., a virtual machine, application, or some other process) operating on a computer system that is separate from the first storage system and the second storage system. That data is stored according to data subsets that are each exclusive of each other. As each data set is migrated, access to the data set by the process is precluded, but access for the remaining data sets is not affected. Once the data migration is complete, access to the data set by the process is restored.

Abstract: A communication protocol system is provided for reliable transport of packets. Transport of packets includes transmitting, by a sender entity over a connection to a receiver entity, a plurality of packets in a first order, maintaining, by the sender entity, one or more sliding windows including a plurality of bits, wherein each bit of the sliding window represents a respective packet of the plurality of packets, receiving, by the sender entity, one or more acknowledgments indicating that one or more of the plurality of packets have been received by the receiver entity, each of the acknowledgments referencing a respective packet of the plurality of packets and modifying, by the sender entity, values of one or more of the plurality of bits in the sliding window corresponding to the one or more acknowledgments received.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing fundus images using fundus image processing machine learning models.

Abstract: Methods, systems, and media for providing media guidance with contextual controls are provided. In some embodiments, the method comprises: causing a user interface to be presented that includes media interface elements, wherein each of the media interface elements is a selectable object that represents at least one media content item; determining that a media interface element from the media interface elements has been selected; in response to determining that the media interface element has been selected, causing the at least one media content item corresponding to the selected media interface element to be presented in a portion of the user interface; and concurrently with causing the at least one media content item to be presented, causing a control interface to be presented that is associated with the selected media interface element.

Abstract: A transformer-transducer model for unifying streaming and non-streaming speech recognition includes an audio encoder, a label encoder, and a joint network. The audio encoder receives a sequence of acoustic frames, and generates, at each of a plurality of time steps, a higher order feature representation for a corresponding acoustic frame. The label encoder receives a sequence of non-blank symbols output by a final softmax layer, and generates, at each of the plurality of time steps, a dense representation. The joint network receives the higher order feature representation and the dense representation at each of the plurality of time steps, and generates a probability distribution over possible speech recognition hypothesis. The audio encoder of the model further includes a neural network having an initial stack of transformer layers trained with zero look ahead audio context, and a final stack of transformer layers trained with a variable look ahead audio context.

Abstract: Operations of a method include receiving a configuration request requesting configuration management of a base configuration package that describes infrastructure of an application. The operations include determining, for the base configuration package, a plurality of package functions. Each package function of the plurality of package functions extends functionality of the base configuration package. The operations include determining, for the base configuration package and the plurality of package functions, a change-proposal function. The change-proposal function proposes one or more of the plurality of package functions to extend the functionality of the base configuration package. The operations include transmitting, to a user device, the change-proposal function and receiving, from the user device, approval of the change-proposal function.

Abstract: An electronic device stores a voice assistant library for execution on the electronic device based on the electronic device having a first device type. The electronic device receives a verbal input from a user. It extracts request information from the verbal input by processing the verbal input using the voice assistant library executing on the device. It transmits a request to a remote system. The electronic device receives a response to the request. The response is generated by the remote system. The electronic device performs an operation in accordance with the response by one or more voice-processing modules of the configured voice assistant library.

Abstract: This document describes an image-capturing doorbell device. In aspects, the image-capturing doorbell device provides a compact, space-efficient, battery-powered, doorbell camera. The architecture of the image-capturing doorbell device is optimized by concentrating sensors at one end of the device and user input mechanism(s) at the opposing end of the device and including a thin and narrow middle portion between the two opposing ends. The sensors include an image sensor and a PIR sensor mounted to the same PCB for space conservation. A camera lens protrudes from an outer surface of an IR window aligned with IR LEDs to mitigate IR flare. The PIR sensor is aligned with a lens that enhances radial motion detection by implementing two stacked rows of lenslets. The user input mechanism includes a light ring formed via a two-shot molding technique with a button to bond the light ring to the button for enhanced waterproofing.

Abstract: This document describes techniques for radio frequency (RF) based micro-motion tracking. These techniques enable even millimeter-scale hand motions to be tracked. To do so, radar signals are used from radar systems that, with conventional techniques, would only permit resolutions of a centimeter or more.

Abstract: This document describes methods, devices, systems, and means for beam failure recovery for wireless communication in an active coordination set (ACS) by a user equipment (UE) in which the UE receives a beam-failure-recovery (BFR) Random Access Channel (RACH) configuration including multiple candidate beam configurations, each candidate beam configuration comprising a candidate BFR sub-beam configuration for each base station in the ACS. The UE detects a beam failure with the ACS and determines a respective link-quality metric for each of the received candidate beam configurations in the BFR RACH configuration. Based on the determined link-quality metrics, the UE selects a candidate beam to use for the wireless communication, and transmits a RACH message that includes an indication of the selected candidate beam, the transmitting being effective to direct the base stations in the ACS to use the selected candidate beam for the wireless communication.

Abstract: Techniques and apparatuses are described for adaptive phase-changing device sharing and handover. In aspects, a second base station shares an adaptive phase-changing device (APD), with a first base station. The second base station determines to use an APD in a communication path for a second wireless link with a second user equipment. The second base station identifies a first base station using the APD in a communication path for a first wireless link with a first UE. Based on identifying the first base station using the APD, the second base station obtains apportioned access to the APD and configures a surface of the APD based on the apportioned access. The second base station uses the APD in a communication path for the second wireless link with the second UE based on the apportioned access.

Abstract: Systems, methods, and media for setting up a media content streaming device are provided. In some implementations, the systems comprise: a first communication interface that is configured to provide a media output to a media content presentation device; a second communication interface that is configured to transfer data; and a hardware processor that is configured to: retrieve a setup application in response to detecting a connection between the media content streaming device and a user device via the second communication interface; cause the setup application to be launched on the user device; receive, from the user device, information relating to setting up the media content streaming device; set up the media content streaming device based on the received information relating to setting up the media content streaming device; and prompt a user to connect the media content streaming device to the media content presentation device via the second communication interface.

Abstract: Systems, methods, and machine-readable media facilitate voice distribution. A wireless, bidirectional audio link with each auxiliary device of a plurality of auxiliary devices may be established and simultaneously managed. A first audio signal comprising a first vocal message may be received as input. A third audio signal may be simultaneously received with receipt of the first audio signal. The third audio signal may include a third vocal message that is different from the first vocal message and a second vocal message. The first audio signal may be buffered with a first input buffer of a control device while the third audio signal is buffered with a second input buffer of the control device. Simultaneous bidirectional translation of the first vocal message and the third vocal message may be performed. The second audio signal including a second vocal message may be provided as output.

Abstract: In accordance with a method for presenting home screen shortcuts, a computer system is coupled to a plurality of media content providers and obtains a respective media application for each of the media content providers. Content is provided for display on the television screen from a first media content provider via a first media application. The computer system obtains a plurality of shortcuts corresponding to a home screen user interface, and the plurality of shortcuts includes one or more media shortcuts and a search shortcut. The home screen user interface is displayed concurrently with the provided content on the television screen, and the one or more media shortcuts and the search shortcut are concurrently displayed on the home screen user interface. In response to selection of the search shortcut, the computer system initiates a shortcut search for one or more additional shortcuts for display on the home screen user interface.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for converting unstructured documents to structured key-value pairs. In one aspect, a method includes: providing an image of a document to a detection model, wherein: the detection model is configured to process the image to generate an output that defines one or more bounding boxes generated for the image; and each bounding box generated for the image is predicted to enclose a key-value pair including key textual data and value textual data, wherein the key textual data defines a label that characterizes the value textual data; and for each of the one or more bounding boxes generated for the image: identifying textual data enclosed by the bounding box using an optical character recognition technique; and determining whether the textual data enclosed by the bounding box defines a key-value pair.

Abstract: An actuator module includes a base plate extending in a plane, a voice coil connected to the base plate, and a magnet assembly that includes a back side facing the base plate and a front side facing away from the base plate. The magnet assembly includes a base layer and sidewalls defining a cup and an inner element including a center magnet mounted within the cup. The sidewalls include a first and second pair of sidewalls. The actuator module includes a rigid frame attached to the base plate, the rigid frame including four stubs. The actuator module also includes a plurality of springs suspending the magnet assembly relative to the frame and base plate, the plurality of springs including a first spring attached to the frame at a first pair of the four stubs and a second spring attached to the frame at a second pair of the four stubs.