Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for storing semi-structured data. One of the methods includes maintaining a plurality of schemas; receiving a first semi-structured data item; determining that the first semi-structured data item does not match any of the schemas in the plurality of schemas; and in response to determining that the first semi-structured data item does not match any of the schemas in the plurality of schemas: generating a new schema, encoding the first semi-structured data item in the first data format to generate the first new encoded data item in accordance with the new schema, storing the first new encoded data item in the data item repository, and associating the first new encoded data item with the new schema.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting utterances of a key phrase in an audio signal. One of the methods includes receiving, by a key phrase spotting system, an audio signal encoding one or more utterances; while continuing to receive the audio signal, generating, by the key phrase spotting system, an attention output using an attention mechanism that is configured to compute the attention output based on a series of encodings generated by an encoder comprising one or more neural network layers; generating, by the key phrase spotting system and using attention output, output that indicates whether the audio signal likely encodes the key phrase; and providing, by the key phrase spotting system, the output that indicates whether the audio signal likely encodes the key phrase.

Abstract: A method includes obtaining a query to create a matrix factorization machine learning model based on a set of training data and determining a model vector and a data vector based on the set of training data. The method also includes determining a dot product between the model vector and the data vector, determining matrices based on the dot product, and generating item vectors using a linear solver based on the matrices. The method also includes generating the matrix factorization machine learning model based on the item vectors and executing the matrix factorization machine learning model.

Abstract: A method of phrase extraction for ASR models includes obtaining audio data characterizing an utterance and a corresponding ground-truth transcription of the utterance and modifying the audio data to obfuscate a particular phrase recited in the utterance. The method also includes processing, using a trained ASR model, the modified audio data to generate a predicted transcription of the utterance, and determining whether the predicted transcription includes the particular phrase by comparing the predicted transcription of the utterance to the ground-truth transcription of the utterance. When the predicted transcription includes the particular phrase, the method includes generating an output indicating that the trained ASR model leaked the particular phrase from a training data set used to train the ASR model.

Abstract: A data parallel pipeline may specify multiple parallel data objects that contain multiple elements and multiple parallel operations that operate on the parallel data objects. Based on the data parallel pipeline, a dataflow graph of deferred parallel data objects and deferred parallel operations corresponding to the data parallel pipeline may be generated and one or more graph transformations may be applied to the dataflow graph to generate a revised dataflow graph that includes one or more of the deferred parallel data objects and deferred, combined parallel data operations. The deferred, combined parallel operations may be executed to produce materialized parallel data objects corresponding to the deferred parallel data objects.

Abstract: A method of performing speech recognition using a two-pass deliberation architecture includes receiving a first-pass hypothesis and an encoded acoustic frame and encoding the first-pass hypothesis at a hypothesis encoder. The first-pass hypothesis is generated by a recurrent neural network (RNN) decoder model for the encoded acoustic frame. The method also includes generating, using a first attention mechanism attending to the encoded acoustic frame, a first context vector, and generating, using a second attention mechanism attending to the encoded first-pass hypothesis, a second context vector. The method also includes decoding the first context vector and the second context vector at a context vector decoder to form a second-pass hypothesis.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, to verify whether an interaction with a particular content displayed on a content platform is an interaction by an actual human user. A request to generate a click verification token can be received from a first application executing on a client device. The request can include data for a first set of parameters for a first interaction and a first message authentication code (MAC) set. If the first MAC set matches a second MAC set, which is generated using some of the data included in the request, the click verification token can be generated. The click verification token can be provided to the first application, which in turn can provide this token to a content verification system. The content verification can use this token to verify whether the first interaction is an actual human user interaction.

Abstract: This document describes techniques and devices for handling an attempt to resume a wireless connection using a base station that supports a different core-network type. While in a current resource control state, such as an inactive state 332, a user equipment (UE) 111 or 112 selects a base station 124 with a different core-network type than a previously selected base station 122 or 121. In some cases, this selected base station 124 does not support the current resource control state or does not enable the UE 111 or 112 to transition to a different resource control state, such as a connected state 330. Consequently, if the UE 111 or 112 attempts to perform a procedure that would result in the UE 111 or 112 transitioning to the unsupported resource control state, the procedure may fail and consequently waste network resources or delay communications with the UE 111 or 112.

Abstract: The present disclosure describes various aspects of complementary 2(N)-bit redundancy for single event upset (SEU) prevention. In some aspects, an integrated circuit includes a data storage element to store a data value, another data storage element to store a complementary data value, a multi-bit data storage element (e.g., a 2-bit storage element) to store both the data value and the complementary data value, and voting logic that may enable a complementary data storage scheme with inter-circuit redundancy to prevent SEU. Additionally, the voting logic of the integrated circuit may enable detection and correction of data value errors and/or enable programming of voting logic criteria, which may be implemented dynamically based on a type of SEU failures that are detected or corrected.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for channel-wise autoregressive entropy models. In one aspect, a method includes processing data using a first encoder neural network to generate a latent representation of the data. The latent representation of data is processed by a quantizer and a second encoder neural network to generate a quantized latent representation of data and a latent representation of an entropy model. The latent representation of data is further processed into a plurality of slices of quantized latent representations of data wherein the slices are arranged in an ordinal sequence. A hyperprior processing network generates a hyperprior parameters and a compressed representation of the hyperprior parameters. For each slice, a corresponding compressed representation is generated using a corresponding slice processing network wherein a combination of the compressed representations form a compressed representation of the data.

Abstract: A method of assembling a display panel includes laminating a back plate to a display layer to form an untrimmed display panel, the back plate including a metal layer that includes a trimming path defined by one or more line segments having reduced metal content compared to other portions of the metal layer. The method further includes trimming the untrimmed display panel along the one or more line segments to define one or more edges of the display panel. For one or more locations along each of the one or more edges defined by the line segments, the metal layer is flush with the corresponding edge of the display panel.

Abstract: A method of interpreting tabular data includes receiving, at a deep tabular data learning network (TabNet) executing on data processing hardware, a set of features. For each of multiple sequential processing steps, the method also includes: selecting, using a sparse mask of the TabNet, a subset of relevant features of the set of features; processing using a feature transformer of the TabNet, the subset of relevant features to generate a decision step output and information for a next processing step in the multiple sequential processing steps; and providing the information to the next processing step. The method also includes determining a final decision output by aggregating the decision step outputs generated for the multiple sequential processing steps.

Abstract: A method including receiving, by a companion device from an optical display, a distortion information associated with a geometric distortion associated with rendering an image on a display of the optical display, distorting, by the companion device, the image using the distortion information, preprocessing, by the companion device, the distorted image based on compression artifacts, compressing, by the companion device, the distorted image, and communicating, by the companion device, the compressed image to the optical display

Abstract: Systems, device and techniques are disclosed for implementing a security configuration change based on one or more base events and a current security configuration. An inference module may identify a security configuration change based on receiving base events from a state storage/event listener and analyzing the base events to determine if a current security configuration is optimal given the base events.

Abstract: Methods, systems, and media for identifying abusive content items are provided.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition are disclosed. In one aspect, a method includes receiving a candidate adword from an advertiser. The method further includes generating a score for the candidate adword based on a likelihood of a speech recognizer generating, based on an utterance of the candidate adword, a transcription that includes a word that is associated with an expected pronunciation of the candidate adword. The method further includes classifying, based at least on the score, the candidate adword as an appropriate adword for use in a bidding process for advertisements that are selected based on a transcription of a speech query or as not an appropriate adword for use in the bidding process for advertisements that are selected based on the transcription of the speech query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, are described for performing asymmetric data communication at a host-device interface of a system. The methods include identifying devices coupled to a host of the system and generating a system topology that identifies a connectivity of the devices and identifies bus lanes that enable data transfers at the system. The host determines that a first connection between the host and a first device of the multiple devices has an asymmetric bandwidth requirement. The host configures a set of bus lanes of a data bus connecting the first device and the host to allocate a different number of the bus lanes to data egress from the host than to data ingress to the host. The bus lanes are configured to allocate the differing number of bus lanes based on the asymmetric bandwidth requirement of the first connection.

Abstract: A method for oversubscribing a host memory of a host running a virtual machine monitor (VMM), comprising, examining a virtual machine (VM) memory for a VM for metadata associated with the VM memory, the metadata maintained by a guest OS running on the VM, collecting the metadata for the VM memory, and managing the VM memory using the metadata for oversubscribing a host memory.

Abstract: Methods, systems, and media for associating multiple users with a media presentation device are provided. In some embodiments, the method comprises: receiving first user preferences; receiving second user preferences; receiving requests to associate the first user preferences and the second user preferences, respectively, with a media presentation device; causing the first user preferences and the second user preferences to be associated with identifying information of the media presentation device; receiving, from the media presentation device, a request to present a slideshow of images based on user preferences and identifying information of the first user device and a third user device; determining that the first user preferences are associated with the media presentation device; determining that user preferences of the third user device are not associated with the media presentation device; and causing content based on the first user preferences to be presented by the media presentation device.

Abstract: Implementations relate to providing information items for display during a communication session. In some implementations, a computer-implemented method includes receiving, during a communication session between a first computing device and a second computing device, first media content from the communication session. The method further includes determining a first information item for display in the communication session based at least in part on the first media content. The method further includes sending a first command to at least one of the first computing device and the second computing device to display the first information item.