Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer-readable media, for speech recognition using multi-dialect and multilingual models. In some implementations, audio data indicating audio characteristics of an utterance is received. Input features determined based on the audio data are provided to a speech recognition model that has been trained to output score indicating the likelihood of linguistic units for each of multiple different language or dialects. The speech recognition model can be one that has been trained using cluster adaptive training. Output that the speech recognition model generated in response to receiving the input features determined based on the audio data is received. A transcription of the utterance generated based on the output of the speech recognition model is provided.

Abstract: Image data is processed for noise reduction before encoding and subsequent decoding. For an input image in a spatial domain, two-dimensional (2-D) wavelet coefficients at multiple levels are generated. Each level includes multiple subbands, each associated with a respective subband type in a wavelet domain. For respective levels, a flat region of a subband is identified, which flat region includes blocks of the subband having a variance no higher than a first threshold variance. A flat block set for the subband type associated with the subband is identified, which includes blocks common to respective flat regions of the subband. A second threshold variance is determined using variances of the flat block set, and is then used for thresholding at least some of the 2-D wavelet coefficients to remove noise. After thresholding, a denoised image is generated in the spatial domain using the levels.

Abstract: A method for conserving resources in a distributed system includes receiving an event-criteria list from a resource controller. The event-criteria list includes one or more events watched by the resource controller and the resource controller controls at least one target resource and is configured to respond to events from the event-criteria list that occur. The method also includes determining whether the resource controller is idle. When the resource controller is idle, the method includes terminating the resource controller, determining whether any event from the event-criteria list occurs after terminating the resource controller, and, when at least one event from the event-criteria list occurs after terminating the resource controller, recreating the resource controller.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for characterizing a gaze position of a user in a query image. One of the methods includes obtaining a query image of a user captured by a camera of a mobile device; obtaining device characteristics data specifying (ii) characteristics of the mobile device, (ii) characteristics of the camera of the mobile device, or (iii) both; and processing a neural network input comprising (i) one or more images derived from the query image and (ii) the device characteristics data using a gaze prediction neural network, wherein the gaze prediction neural network is configured to, at run time and after the gaze prediction neural network has been trained, process the neural network input to generate a neural network output that characterizes a gaze position of the user in the query image.

Abstract: A method includes receiving a sequence of acoustic frames extracted from audio data corresponding to an utterance. During a first pass, the method includes processing the sequence of acoustic frames to generate N candidate hypotheses for the utterance. During a second pass, and for each candidate hypothesis, the method includes: generating a respective un-normalized likelihood score; generating a respective external language model score; generating a standalone score that models prior statistics of the corresponding candidate hypothesis; and generating a respective overall score for the candidate hypothesis based on the un-normalized likelihood score, the external language model score, and the standalone score. The method also includes selecting the candidate hypothesis having the highest respective overall score from among the N candidate hypotheses as a final transcription of the utterance.

Abstract: Embodiments according to examples aspects of the present disclosure provide for an example computer-implemented method. The example method can include obtaining a data structure configured for rendering a plurality of content containers on a user device, the plurality of content containers providing a collapsed configuration and an expanded configuration. The example method can include rendering, based on a first rate parameter descriptive of a user input associated with the user device, one or more of the plurality of content containers according to the collapsed configuration. The example method can include rendering, based on a second rate parameter, at least one of the plurality of content containers according to the expanded configuration.

Abstract: The technology described in this document can be embodied in a computer-implemented method that includes receiving, at a processing system, a first signal including an output of a speaker device and an additional audio signal. The method also includes determining, by the processing system, based at least in part on a model trained to identify the output of the speaker device, that the additional audio signal corresponds to an utterance of a user. The method further includes initiating a reduction in an audio output level of the speaker device based on determining that the additional audio signal corresponds to the utterance of the user.

Abstract: Systems and methods of offloading multicast virtual network packet processing to a network interface card are provided. In an example implementation, a network interface card can route packets in a virtual network. The network interface card can be configured to receive a data packet having a multicast header for transmission to a plurality of destination virtual machines. The network interface card can retrieve a list of next hop destinations for the data packet. The network interface card can replicate the packet for each next hop destination. The network interface card can encapsulate each replicated packet with a unicast header that includes a next hop destination virtual IP address indicating the next hop destination and a source virtual IP address, and transmit the encapsulated packets.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for an automated calling system are disclosed. In one aspect, a method includes the actions of receiving audio data of an utterance spoken by a user who is having a telephone conversation with a bot. The actions further include determining a context of the telephone conversation. The actions further include determining a user intent of a first previous portion of the telephone conversation spoken by the user and a bot intent of a second previous portion of the telephone conversation outputted by a speech synthesizer of the bot. The actions further include, based on the audio data of the utterance, the context of the telephone conversation, the user intent, and the bot intent, generating synthesized speech of a reply by the bot to the utterance. The actions further include, providing, for output, the synthesized speech.

Abstract: Methods, systems, and media for secure authentication of users using one or more biometric recognition systems are provided. In some embodiments, the method comprises: receiving an indication that a biometric identifier is to be used to authenticate a user to a service; receiving (i) the biometric identifier of the user from a capture device and (ii) knowledge-based secondary information associated with the user from an input device; determining a Voronoi cell identifier that corresponds to the biometric identifier; calculating a hash of the Voronoi cell identifier and the knowledge-based secondary information; transmitting the hash to a server device for verification; in response to transmitting the hash to the server device, receiving a response indicating whether the hash matches a previously stored hash that was stored in the server device; and determining whether to automatically authenticate the user to the service based on the response from the server device.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword detection on multiple devices are disclosed. In one aspect, a method includes the actions of receiving, by a first computing device, audio data that corresponds to an utterance. The actions further include determining a first value corresponding to a likelihood that the utterance includes a hotword. The actions further include receiving a second value corresponding to a likelihood that the utterance includes the hotword, the second value being determined by a second computing device. The actions further include comparing the first value and the second value. The actions further include based on comparing the first value to the second value, initiating speech recognition processing on the audio data.

Abstract: Techniques and apparatuses are described for user equipment-coordination set, UECS, hybrid automatic repeat request, HARQ, that establish a HARQ timeline that is specific to the capabilities of a respective UECS. Compared to a single user equipment, UE, the HARQ timeline for a UECS depends on a number of factors, such as the joint processing capability in the UECS, latency of communication over a local wireless network between the UEs in the UECS, or the like. Based on its capabilities, the UECS can request uplink and/or downlink processing delay times or a UECS-specific HARQ timeline from a base station. The base station grants the uplink and/or downlink processing delay times or the UECS-specific HARQ timeline to the UECS in a layer-1, layer-2, or a layer-3 control message. The use of a UECS-specific HARQ timeline increases the reliability of HARQ signaling for uplink and downlink communication between the UECS and a base station.

Abstract: Processing of acoustic features of audio data to generate one or more revised versions of the acoustic features, where each of the revised versions of the acoustic features isolates one or more utterances of a single respective human speaker. Various implementations generate the acoustic features by processing audio data using portion(s) of an automatic speech recognition system. Various implementations generate the revised acoustic features by processing the acoustic features using a mask generated by processing the acoustic features and a speaker embedding for the single human speaker using a trained voice filter model. Output generated over the trained voice filter model is processed using the automatic speech recognition system to generate a predicted text representation of the utterance(s) of the single human speaker without reconstructing the audio data.

Abstract: Implementations are directed to updating a trained voice bot that is deployed for conducting conversations on behalf of a third-party. A third-party developer can interact with a voice bot development system that enables the third-party developer to train, update, validate, and monitor performance of the trained voice bot. In various implementations, the trained voice bot can be updated by updating a corpus of training instances that was initially utilized to train the voice bot, and updating the trained voice bot based on the updated corpus. In some implementations, the corpus of training instances may be updated in response to identifying occurrence(s) of behavioral error(s) of the trained voice bot while the conversations are being conducted on behalf of the third-party. In additional or alternative implementations, the corpus of training instances may be updated in response to determining the trained voice bot does not include a desired behavior.

Abstract: A computer-implemented method includes receiving a batch of neural network inputs to be processed using a neural network on a hardware circuit. The neural network has multiple layers arranged in a directed graph and each layer has a respective set of parameters. The method includes determining a partitioning of the neural network layers into a sequence of superlayers. Each superlayer is a partition of the directed graph that includes one or more layers. The method includes processing the batch of inputs using the hardware circuit, which includes, for each superlayer in the sequence: i) loading the respective set of parameters for the layers in the superlayer into memory of the hardware circuit, and ii) for each input in the batch, processing the input through each of the layers in the superlayer using the parameters in the memory of the hardware circuit to generate a superlayer output for the input.

Abstract: A device for generating a qubit control signal includes: a first signal envelope generator circuit including a first multiple of signal sources, in which an output of each signal source of the first multiple of signal sources is combined to provide a first cumulative output; and a first mixer circuit coupled to the first signal envelope generator circuit, in which the first cumulative output is coupled to a first input of the first mixer circuit, and an output of the first mixer circuit includes a first qubit control signal.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a machine learning task on a network input to generate a network output. In one aspect, one of the systems includes an attention neural network configured to perform the machine learning task, the attention neural network including one or more attention layers, each attention layer comprising an attention sub-layer and a feed-forward sub-layer that applies an element-wise multiplication between two vectors generated as a result of two different linear transformations performed on the same attended layer input.

Abstract: An example method includes displaying, by a graphical user interface of a computing device, an image comprising a target region. The target region may be smaller than an entirety of the image. The method includes providing, by the graphical user interface, a user-adjustable control to adjust a desired local brightness exposure level for the target region. The method includes receiving, by the user-adjustable control, a user indication of the desired local brightness exposure level for the target region. The method includes adjusting the local brightness exposure level for the target region in the image in response to the user indication.

Abstract: A navigation system can identify locations of interest at a route destination. Those locations of interest at the destination can fall within a predetermined distance of the destination or those that are responsive to a query that includes the destination as a parameter. The navigation system can receive a selection of an identified location of interest near the destination, and update a route to terminate at the identified location of interest rather than the original destination. Information identifying the origin and destination for the route can be collected at a first user device and the route can be displayed at a second user device. Additional information is then received from the first user device (such as a selection of an identified location of interest near the destination). This additional information received at the first user device is used to update the route that is displayed on the second user device.

Abstract: Methods, systems, and apparatus include computer programs encoded on a computer-readable storage medium, including a method for providing content. Search results responsive to a query are identified including a first search result in a top set of search results, the first search result associated with a first entity. A first eligible content item is identified for presentation along with the search results, the first eligible content item associated with the first entity. A combined content item is identified that is a combination of the first search result and first eligible content item and is to be presented as a search result responsive to the query. The combined content item is augmented including: identifying entities related to the first entity, identifying content items that are associated with the related entities, selecting at least one identified content item, and using content from the selected content items to augment the combined content item.