Abstract: Systems, methods and computer-readable storage media utilized to train a machine-learning architecture. One method includes receiving, by one or more processing circuits, a data set. The method further includes determining, by the one or more processing circuits, a first portion of the data set associated with a plurality of entities. The method further includes training, by the one or more processing circuits and utilizing the first portion of the data set, an entity model. The method further includes determining, by the one or more processing circuits, a second portion of the data set associated with a first subset of entities and determining a second subset of entities. The method further includes freezing, by the one or more processing circuits, one or more parameters associated with the second subset of entities and training, utilizing the second portion of the data set, the entity model.

Abstract: Mitigating the reality gap through training and utilization of at least one difference model. The difference model can be utilized to generate, for each of a plurality of instances of simulated state data generated by a robotic simulator, a corresponding instance of modified simulated state data. The difference model is trained so that a generated modified instance of simulated state data is closer to “real world data” than is a corresponding initial instance of simulated state data. Accordingly, the difference model can be utilized to mitigate the reality gap through modification of initially generated simulated state data, to make it more accurately reflect what would occur in a real environment. Moreover, the difference representation from the difference model can be used as input to the control policy to adapt the control learned from simulator to the real environment.

Abstract: Systems and methods for determining whether to combine responses from multiple automated assistants. An automated assistant may be invoked by a user utterance, followed by a query, which is provided to a plurality of automated assistants. A first response is received from a first automated assistant and a second response is received from a second automated assistant. Based on similarity between the responses, a primary automated assistant determines whether to combine the responses into a combined response. Once the combined response has been generated, one or more actions are performed in response to the combined response.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining, for each of multiple words or sub-words, audio data corresponding to multiple users speaking the word or sub-word; training, for each of the multiple words or sub-words, a pre-computed hotword model for the word or sub-word based on the audio data for the word or sub-word; receiving a candidate hotword from a computing device; identifying one or more pre-computed hotword models that correspond to the candidate hotword; and providing the identified, pre-computed hotword models to the computing device.

Abstract: A method includes obtaining an utterance from a user including a user query directed toward a digital assistant. The method includes generating, using a language model, a first prediction string based on the utterance and determining whether the first prediction string includes an application programming interface (API) call to invoke a program via an API. When the first prediction string includes the API call to invoke the program, the method includes calling, using the API call, the program via the API to retrieve a program result; receiving, via the API, the program result; updating a conversational context with the program result that includes the utterance; and generating, using the language model, a second prediction string based on the updated conversational context. When the first prediction string does not include the API call, the method includes providing an utterance response to the utterance based on the first prediction string.

Abstract: Methods, devices, systems, and means for intra-UECS communication by a coordinating user equipment, UE, in a user equipment-coordination set, UECS, are described herein. The coordinating UE allocates first air interface resources to a second UE and second air interface resources to a third UE for intra-UECS communication. The coordinating UE receives, using the allocated first air interface resources, an Internet Protocol, IP, data packet from the second UE in the UECS. The coordinating UE determines that a destination address included in the IP data packet is an address of the third UE and transmits, using the allocated second air interface resources, the IP data packet to the third UE.

Abstract: A joint segmenting and ASR model includes an encoder to receive a sequence of acoustic frames and generate, at each of a plurality of output steps, a higher order feature representation for a corresponding acoustic frame. The model also includes a decoder to generate based on the higher order feature representation at each of the plurality of output steps a probability distribution over possible speech recognition hypotheses, and an indication of whether the corresponding output step corresponds to an end of segment (EOS).

Abstract: A method includes receiving user queries directed toward a cloud-based assistant service. For each received user query directed toward the cloud-based assistant service, the method also includes extracting one or more attributes from the user query and logging the user query into one or more of a plurality of category buckets based on the one or more attributes extracted from the user query. The method also includes determining when at least one of the plurality of category buckets includes a threshold number of the user queries logged into the at least one category bucket, and when the at least one of the plurality of category buckets includes the threshold number of the user queries, generating a distilled model of the cloud-based assistant service. The distilled model of the cloud-based assistant service is configured to execute on one or more target client devices.

Abstract: A method includes obtaining a key status for a first cryptographic key. The first cryptographic key is used to encrypt replicated data of a first replication instance. The method also includes determining, based on the key status, that the first cryptographic key is inaccessible which causes the first replication instance to be unavailable. In response to determining that the first cryptographic key is inaccessible, the method includes scheduling a second replication instance to be unavailable after a threshold amount of time has passed. The second replication instance includes replicated data encrypted by a second cryptographic key that is accessible. After the threshold amount of time has passed and when the first cryptographic key is still inaccessible, the method includes setting the second replication instance as unavailable.

Abstract: This document describes systems and techniques directed at enlarging active areas of displays in electronic devices. In aspects, a display includes a grid of transistors positioned within a display panel module to control an illumination of one or more electroluminescent layers. Routing lines extend from one or more transistors of the grid of transistors to at least one electroluminescent layer. In this way, the at least one electroluminescent layer can be positioned away from the grid of transistors and disposed above portions of display panel module driving circuitry. As a result, active areas of displays can be enlarged and information content can be maximized without a panel border area allotted to the display panel module driving circuitry surrounding transistors having to be reduced.

Abstract: A method of training an accent recognition model includes receiving a corpus of training utterances spoken across various accents, each training utterance in the corpus including training audio features characterizing the training utterance, and executing a training process to train the accent recognition model on the corpus of training utterances to teach the accent recognition model to learn how to predict accent representations from the training audio features. The accent recognition model includes one or more strided convolution layers, a stack of multi-headed attention layers, and a pooling layer configured to generate a corresponding accent representation.

Abstract: This document describes systems and techniques directed at enlarging active areas of displays using variable pixel and/or transistor densities. In aspects, a display includes a cover layer positioned as a topmost layer and an array of pixels positioned thereunder. A plurality of transistors, positioned under the array of pixels, may control an electrical activation of one or more pixels within the array of pixels. In implementations, the plurality of transistors define a smaller area than the array of pixels such that at least one pixel of the array of pixels extends beyond the area defined by the plurality of transistors and above driving circuitry. Variable pixel and/or transistor densities can support the enlarged active area of displays and improve user experience.

Abstract: A method includes obtaining a stream of consecutive network configuration snapshots each including network configuration information. The method also includes determining that first network configuration information of a first network configuration snapshot of the network from the stream of consecutive network configuration snapshots for the network is not the same as second network configuration information of a second network configuration snapshot of the network from the stream of consecutive network configuration snapshots for the network. The method also includes generating a reachability differentiation graph that identifies a net change to reachability from the first network configuration information and the second network configuration information based on determining that the first network configuration information is not the same as the second network configuration information.

Abstract: A method includes receiving training data including a corpus of multilingual unspoken textual utterances, a corpus of multilingual un-transcribed non-synthetic speech utterances, and a corpus of multilingual transcribed non-synthetic speech utterances. For each un-transcribed non-synthetic speech utterance, the method includes generating a target quantized vector token and a target token index, generating contrastive context vectors from corresponding masked audio features, and deriving a contrastive loss term. The method also includes generating an alignment output, generating a first probability distribution over possible speech recognition hypotheses for the alignment output, and determining an alignment output loss term. The method also includes generating a second probability distribution over possible speech recognition hypotheses and determining a non-synthetic speech loss term.

Abstract: A method of training a language model for rare-word speech recognition includes obtaining a set of training text samples, and obtaining a set of training utterances used for training a speech recognition model. Each training utterance in the plurality of training utterances includes audio data corresponding to an utterance and a corresponding transcription of the utterance. The method also includes applying rare word filtering on the set of training text samples to identify a subset of rare-word training text samples that include words that do not appear in the transcriptions from the set of training utterances or appear in the transcriptions from the set of training utterances less than a threshold number of times. The method further includes training the external language model on the transcriptions from the set of training utterances and the identified subset of rare-word training text samples.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a generative neural network to convert conditioning text inputs to audio outputs using energy scores.

Abstract: Various devices, systems and methods for performing contactless monitoring of the sleep of multiple users over a same time period are presented herein. Clustering may be performed based on data received from a radar sensor. Based on the clustering performed on the data received from the radar sensor, a determination may be made that two users are present within the region. In response to determining that two users are present, a midpoint location may be calculated between the clusters. A first portion of the data may be mapped to a first user and a second portion of the data may be mapped to a second user based on the calculated midpoint. Separate sleep analyses may be performed for the first user and the second user.

Abstract: Mechanisms for removing objectionable and/or inappropriate content from media content items are provided. In some embodiments, the method comprises: receiving a first media content item and a dictionary, wherein the first media content item includes an audio component and a video component; identifying a plurality of scenes and a plurality of scene breaks associated with the first media content item; transcribing the audio component of the first media content item to produce transcribed audio; comparing the transcribed audio to entries in the dictionary and storing matches between the transcribed audio and the entries; and generating a second media content item by removing at least a portion of at least one of the audio component and the video component based on the matches.

Abstract: The present disclosure provides systems and methods that generating health diagnostic information from an audio recording. A computing system can include a machine-learned health model comprising that includes a sound model trained to receive data descriptive of a patient audio recording and output sound description data. The computing system can include a diagnostic model trained to receive the sound description data and output a diagnostic score. The computing system can include at least one tangible, non-transitory computer-readable medium that stores instructions that, when executed, cause the processor to perform operations. The operations can include obtaining the patient audio recording; inputting data descriptive of the patient audio recording into the sound model; receiving, as an output of the sound model, the sound description data; inputting the sound description data into the diagnostic model; and receiving, as an output of the diagnostic model, the diagnostic score.