Abstract: Features are disclosed for detecting words in audio using contextual information in addition to automatic speech recognition results. A detection model can be generated and used to determine whether a particular word, such as a keyword or “wake word,” has been uttered. The detection model can operate on features derived from an audio signal, contextual information associated with generation of the audio signal, and the like. In some embodiments, the detection model can be customized for particular users or groups of users based usage patterns associated with the users.

Abstract: The disclosure relates in some aspects to location reporting and paging for satellite communication. In one aspect, the disclosure relates to a user terminal (UT) sending a message to report information about the UT's location. A threshold may be used to control whether (e.g., when) the UT reports its location. The location information may be used to enforce area (e.g., country) restrictions for the UT. In one aspect, the disclosure relates to forwarding paging messages between network access controllers. For example, a network access controller that is not able to page a UT may forward a paging message to another network access controller. In one aspect, the disclosure relates to a list of paging areas that indicates where a UT need not perform a paging area update.

Abstract: Forming a protective coating ex situ in an atomic layer deposition process to coat one or more chamber components subsequently installed in a reaction chamber provides a number of benefits over more conventional coating methods such as in situ deposition of an undercoat. In certain cases the protective coating may have a particular composition such as aluminum oxide, aluminum fluoride, aluminum nitride, yttrium oxide, and/or yttrium fluoride. The protective coating may help reduce contamination on wafers processed using the coated chamber component. Further, the protective coating may act to stabilize the processing conditions within the reaction chamber, thereby achieving very stable/uniform processing results over the course of processing many batches of wafers, and minimizing radical loss. Also described are a number of techniques that may be used to restore the protective coating after the coated chamber component is used to process semiconductor wafers.

Abstract: One variation of a method for modeling mobile advertisement consumption includes: serving a first advertisement in an advertising campaign to a computing device associated with a user, accessing a first set of engagement data, recorded by the first advertisement, representing a first set of interactions between the user and the first advertisement at the computing device; accessing a model linking user interactions with a set of advertisements within the advertising campaign and a target outcome for the advertising campaign; estimating a predicted set of interactions between the user and a second advertisement in the advertising campaign based on the model and the first set of engagement data; and in response to the predicted set of interactions anticipating the target outcome, serving the second advertisement, in the advertising campaign, to the user.

Abstract: A system including: at least one processor; and at least one memory having stored thereon computer program code that, when executed by the at least one processor, controls the system to: receive a data model identification and a dataset; in response to determining that the data model does not contain a hierarchical structure, perform expectation propagation on the dataset to approximate the data model with a hierarchical structure; divide the dataset into a plurality of channels; for each of the plurality of channels: divide the data into a plurality of microbatches; process each microbatch of the plurality of microbatches through parallel iterators; and process the output of the parallel iterators through single-instruction multiple-data (SIMD) layers; and asynchronously merge results of the SIMD layers.

Abstract: A method includes receiving a first image including color data and depth data, determining a viewpoint associated with an augmented reality (AR) and/or virtual reality (VR) display displaying a second image, receiving at least one calibration image including an object in the first image, the object being in a different pose as compared to a pose of the object in the first image, and generating the second image based on the first image, the viewpoint and the at least one calibration image.

Abstract: A technique for repairing an indirect addressing structure of a file system damaged by corruption of a mid-level mapping (MID) page includes scanning selected leaf pages to identify leaf pages associated with the corrupted MID page, then recreating the MID page by recreating pointers to the identified leaf pages. The scanning includes (1) based on an association of groups of leaf pages with corresponding sets of families of storage objects, scanning the leaf pages of only those groups of leaf pages associated with the family of storage objects for the corrupted MID page. The scanning further includes (2) performing a two-pass process including first identifying all leaf pages for the logical offset range of the corrupted MID page and then pruning those identified leaf pages that are reachable via non-corrupted MID pages, yielding the leaf pages for the corrupted MID page only, usable to recreate the corrupted MID page.

Abstract: A method of repairing an indirect addressing structure of a file system damaged by corruption of a virtual data block (VDB) mapping data stored in corresponding physical data blocks (PDBs) includes scanning PDB descriptors to identify PDBs storing data mapped by the corrupted VDB, where each identified PDB includes a set of PDB entries each having a backward pointer identifying a corresponding VDB entry of a corresponding VDB. The identified PDBs are scanned to identify PDB entries whose backward pointers refer to VDB entries of the corrupted VDB, then a replacement VDB is created by (1) for each of the identified PDB entries, recreating a corresponding VDB entry including a forward pointer to the identified PDB entry, and (2) incorporating the recreated VDB entries into the replacement VDB. The replacement VDB is then incorporated into the indirect addressing structure in place of the corrupted VDB.

Abstract: Systems, methods, and computer-readable media for hierarchichal sharding of flows from sensors to collectors. A first collector can receive a first portion of a network flow from a first capturing agent and determine that a second portion of the network flow was not received from the first capturing agent. The first collector can then send the first portion of the network flow to a second collector. A third collector can receive the second portion of the network flow from a second capturing agent and determine that the third collector did not receive the first portion of the network flow. The third collector can then send the second portion of the network flow to the second collector. The second collector can then aggregate the first portion and second portion of the network flow to yield the entire portion of the network flow.

Abstract: A computer-implemented method comprising extracting profile information for a profile, converting the profile information into a readable file format, and storing the converted profile information. The method may include creating a database file from the converted profile information and copying the database file. The method may include updating the database file copy with profile changes, and loading the updated database file copy to generate an updated profile where the updated database file copy is mapped according to metadata in the updated database file copy. The method may include altering a profile pointer reference to point to the updated database file copy to activate the updated profile, wherein activating the at least one updated profile activates the one or more profile changes.

Abstract: A method includes receiving audio data encoding an utterance spoken by a native speaker of a first language, and receiving a biasing term list including one or more terms in a second language different than the first language. The method also includes processing, using a speech recognition model, acoustic features derived from the audio data to generate speech recognition scores for both wordpieces and corresponding phoneme sequences in the first language. The method also includes rescoring the speech recognition scores for the phoneme sequences based on the one or more terms in the biasing term list, and executing, using the speech recognition scores for the wordpieces and the rescored speech recognition scores for the phoneme sequences, a decoding graph to generate a transcription for the utterance.

Abstract: An expandable rack mountable computing device for use in a variety of telecommunications and computing racks comprises a computing device chassis, at least one expandable baffle, at least one biasing mechanism, and at least one expansion mechanism. During installation of the computing device into a rack, the biasing mechanism holds the at least one expandable baffle in a retracted position at least partially inside of the computing device chassis. After installation of the computing device into a rack, the expansion mechanism pushes the at least one expandable baffle into a retracted position at least partially outside of the computing device chassis.

Abstract: An analog-to-digital converter (10) comprises a first and a second sampling capacitor (24, 25), a first integrator (26), a first and a second input switch (31, 32) coupling a first input terminal (11) and a common mode terminal (39) to a first electrode of the first sampling capacitor (24), a third and a fourth input switch (33, 34) coupling a second input terminal (12) and the common mode terminal (39) to a first electrode of the second sampling capacitor (25), a fifth and a sixth input switch (35, 36) coupling a second electrode of the first sampling capacitor (24) to an amplifier common mode terminal (40) and the first integrator input (27), and a seventh and an eighth input switch (37, 38) coupling a second electrode of the second sampling capacitor (25) to the amplifier common mode terminal (40) and the second integrator input (28).

Abstract: A burner for a combustor of a turbomachine includes a pilot nozzle with inline premixing. The pilot nozzle is formed in an aft end of the burner. An air inlet is formed in a forward end of the burner in fluid communication with the pilot nozzle. A mixing channel extends along the axial direction between the air inlet and the pilot nozzle such that the air inlet is in fluid communication with the pilot nozzle via the mixing channel. An annular fuel plenum extends along the circumferential direction. A fuel port is in fluid communication with the annular fuel plenum and the mixing channel, the fuel port includes an outlet configured to inject fuel into the mixing channel such that a shear flow is induced.

Abstract: Methods and systems for forming a fleet and positioning vehicles in the fleet are disclosed. The system comprises a data transmission interface and a data processing unit. The data processing unit sends data to and receives data from the data transmission interface. The data processing unit determines an itinerary of each of a group of vehicles requesting to join a fleet. The data processing unit compares the itinerary of all vehicles of the group of vehicles. The data processing unit determines a second vehicle to form a fleet with a first vehicle from the group of vehicles. The data transmission interface communicates to the first vehicle a proposal to form a fleet with the second vehicle. The data processing unit determines if the first vehicle accepts the proposal, and if yes, determines instructions for the first vehicle how to join the fleet.

Abstract: Methods for controlled microstructure creation utilize seeding of amorphous layers prior to annealing. Seed crystals are formed on an amorphous layer or layers. The material, size, and spacing of the seed crystals may be varied, and multiple seed layers and/or amorphous layers may be utilized. Thereafter, the resulting assembly is annealed to generate a crystalline microstructure. Via use of these methods, devices having desirable microstructural properties are enabled.

Abstract: Embodiments of the present disclosure present devices, methods, and computer readable medium for techniques for measuring operational performance metrics, and presenting these metrics through an application programming interface (API) for developers to access for optimizing their applications. Exemplary metrics can include central processing unit or graphics processing unit time, foreground/background time, networking bytes (per application), location activity, display average picture luminance, cellular networking condition, peak memory, number of logical writes, launch and resume time, frame rates, and hang time. Regional markers can also be used to measure specific metrics for in application tasks. The techniques provide multiple user interfaces to help developers recognize the important metrics to optimize the performance of their applications. The data can be normalized over various different devices having different battery size, screen size, and processing requirements.

Abstract: A battery array frame according to an exemplary aspect of the present disclosure includes, among other things, a frame body, and a thermal fin including a body embedded in the frame body and a leg that extends outside of the frame body. The thermal fin is flexible between a first position in which the leg is spaced farther from a surface of the frame body and a second position in which the leg is spaced closer to the surface of the frame body.

Abstract: Methods and systems for reducing the data recovery time when failing back to a primary system from a secondary data backup system using virtual machine conversions and incremental exportation for backed-up virtual machine snapshots are described. Upon detection that the primary system is available to receive replicated snapshots for a virtual machine, the secondary data backup system may identify a first virtual machine snapshot out of a chain of snapshots for the virtual machine (e.g., the earliest in time snapshot within the chain or a middle snapshot in the middle of the chain) and transfer a full image snapshot for the first virtual machine snapshot to the primary system. Thereafter, data changes between the next virtual machine snapshot to be transferred to the primary system and the most-recently hydrated virtual machine on the primary system may be identified and then transferred to the primary system.

Abstract: A system provides notification of a time range for an order delivery. The system identifies estimated maximum delivery dates for each order in historical order data, and generates estimated minimum delivery dates for a first portion of the historical order data using a predictive model, wherein the predictive model is trained on a second portion of the historical order data. The system evaluates accuracy of the predictive model by determining a proportion of the first portion of the historical order data that had an actual delivery date between its corresponding estimated minimum delivery date and its corresponding estimated maximum delivery date, and revises a feature matrix to improve the accuracy of the predictive model.