Abstract: A search query relating to a search for people and including one or more search terms is received. A search graph for the search is created from a previously created graph which is a directional graph in which people are represented by nodes and content is represented by nodes or edges. The search graph is generated by filtering out nodes or edges that represent content that is determined not to be relevant to the search terms in the search query; and, for the remaining content that is determined to be relevant to the search terms in the search query, adding to the graph a measure of the relevance of the content to the search terms. The search results are generated by operating on the search graph to identify people in the search graph who are sources of content that is relevant to the search terms.

Abstract: The techniques described herein improve user interface consistency when presenting a guest application running on a guest operating system as if it were a native application running on a host operating system. Specifically, window placement limitations imposed by the guest operating system are overcome by coordinating size and location between the guest application and a host proxy application. When the proxy application is confined to a single display, the size and location of the guest application and corresponding proxy application are synchronized. When the proxy application is moved or resized so as to be rendered on multiple displays, the application hosting engine decouples the location synchronization, leaving the guest application within a single display. Before being forwarded to the guest application, user interface events received by the proxy application are adjusted according to the different locations of the guest application and the proxy application.

Abstract: Methods and systems are provided for determining the category of a software application utilizing machine learning (ML) and knowledge graph techniques, and for controlling access to the application by a user based on the category and configured time restrictions for the user. The system includes a feature set extractor and a category predictor with a trained ML model. The trained ML model generates the category of the application based on a feature(s) of the application. The generated category is indicated in a data structure. An access request handler receives a request related to access to the application from a user device. A category determiner determines the category of the application from the data structure. A time usage manager determines an available time usage for the category and the specified user. The access arbiter responds to the request from the user device with the available time usage.

Abstract: The disclosed technology is generally directed to the compression of inverted indexes. In one example of the technology, an inverted index that includes a plurality of posting lists and metadata is provided. The inverted index indicates compression settings that are associated with the plurality of posting lists. At periodic scheduled times, a regeneration is performed on the inverted index. The regeneration includes decompressing the inverted index. The decompressing uses the compression settings indicated by the inverted index. The regeneration further includes determining compression settings to use during a next periodic scheduled time of the plurality of periodic scheduled times, such that at least a first posting list of the plurality of posting lists uses a different compression setting than a second posting list of the plurality of posting lists.

Abstract: A rack in a datacenter is powered by a first power feed and a second power feed. The rack supports a plurality of servers which have a maximum combined power consumption which is greater than a maximum supplied power from either the first power feed or the second power feed. When power is lost from one of the power feeds, a rack manager reduces the total power consumption of the plurality of servers by throttling at least one of the servers and/or shutting off at least one of the plurality of servers.

Abstract: The present disclosure relates to utilizing a buffer management system to efficiently manage and deallocate memory buffers utilized by multiple processing roles on computer hardware devices. For example, the buffer management system utilizes distributed decentralized memory buffer monitoring in connection with augmented buffer pointers to deallocate memory buffers accurately and efficiently. In this manner, the buffer management system provides an efficient approach for multiple processing roles to consume source data stored in a memory buffer and to deallocate the buffer only after all roles have finished using it.

Abstract: A computer search service can receive an enterprise graph query from a client computing device that is remote from the computer search service, with the enterprise graph query identifying an actor object set, a target object set, and a relationship between the actor object set and the target object set. The search service can return results of the enterprise graph query to the client device. The search service can receive from the client device an indication of user input instructing the search service to persist the enterprise graph query, with the indication of user input also instructing the search service to associate the persisted enterprise graph query with an entity. Moreover, in response to receiving the indication of user input, the search service can persist the enterprise graph query as well as associate the persisted enterprise graph query with the entity and/or enter a subscription entry for the query.

Abstract: The disclosure herein describes processing data operation requests and performing associated operations on pages of a record data structure. A data operation is generated based on a received data operation request. The data operation request includes a bookkeeping page identifier. A bookkeeping page of the record index structure is identified based on the bookkeeping page identifier and an index page of the record index structure with a parameter of the data operation. If a timestamp in metadata of the identified index page matches a corresponding timestamp associated with the identified index page in metadata of the bookkeeping page, the data operation is performed on the identified index page. The disclosure enables the use and maintenance of large data indexes in data storage systems where data entries expire over time due to automatic temporal compaction processes.

Abstract: Custom source code generation models are generated by tuning a pre-trained deep learning model by freezing the model parameters and optimizing a prefix. The tuning process is distributed across a user space and a model space where the embedding and output layers are performed in the user space and the execution of the model is performed in a model space that is isolated from the user space. The tuning process updates the embeddings of the prefix across the separate execution spaces in a manner that preserves the privacy of the data used in the tuning process.

Abstract: In one example of the technology, device information associated with a device upgrade and a plurality of devices includes risk parameters including values associated with a minimum health value that is associated with a minimum acceptable number of healthy devices among the plurality of devices and a confidence value associated with a minimum acceptable probability that the number of healthy devices among the plurality of devices is at least as great as the minimum health value; and, for each device a success probability value that is associated with a probability that the device will be healthy after the device upgrade is performed on the device. A Poisson binomial distribution is iteratively used to determine a set of devices among the plurality of device for which the largest possible number of devices are included in the set of devices while meeting the risk parameters. The set of devices is then upgraded.

Abstract: A user device having a presentation application includes a processor; a memory storing the presentation application for implementation by the processor, alone or in combination with other processors. The application includes a user interface allowing a user to request integration of multiple slides into a single canvas with a common background generated by an Artificial Intelligence (AI) model based on backgrounds of the multiple slides being integrated into the single canvas.

Abstract: Methods, systems, apparatuses, and computer program products are provided herein for determining the handedness of input provided by a user via a touch interface. For instance, for each touch-based input detected, a score indicating a probability whether the touch-based input was inputted by a particular hand of the user is generated. A classification for the touch-based input is then generated based on a drift diffusion model-based technique in which inter-dependencies between a series of touch-based input are approximated. The determined classifications are used to determine the handedness of the user.

Abstract: Systems and methods for using a generative artificial intelligence (AI) model using a multi-turn process to generate a suggested draft reply to a selected message. A first turn of the multi-turn process uses a shorter prompt including at least a portion of the body of the selected message and that requests multiple draft replies from the AI model. The resulting AI-generated draft replies are shortened, summarized, and/or otherwise converted into a plurality of shortened summaries that are presented as reply options to a user. Upon selecting a shortened summary, a more robust prompt is generated in a second turn with the AI model with the selected reply option to generate a more complex suggested draft reply to the selected message. Additionally, various customization options are provided, which when selected, reframe a query presented to the AI model to generate a more relevant and personalized response.

Abstract: Various techniques for providing perspective annotation to numerical representations are disclosed herein. For example, a method includes detecting a numerical representation in an original content and retrieving one or more perspectives from a database based on the detected numerical representation. The one or more perspectives individually include a restatement of information contained in the numerical representation. The method can also include annotating the original content with the retrieved one or more perspectives to form an annotated content.

Abstract: In embodiments of the present disclosure, there is provided an approach for fact checking based on semantic graphs. According to embodiments of the present disclosure, after obtaining a text to be fact checked, a plurality of evidence sentences related to the text are retrieved from an evidence database. Then, semantic graphs of the text and the evidence sentences are constructed based on the semantic analysis, and a veracity of a statement in the text can be determined based on the semantic graphs. Embodiments of the present disclosure propose a graph-based reasoning approach for fact checking, and use the constructed semantic graphs to facilitate verification of the truthfulness of the text, thereby improving the accuracy for fact checking.

Abstract: A system and method for presenting different graphical representations of a user to different end-users in a network-based communication environment are provided. A first end-user has a first a social distance from a user in a social graph while a second end-user has a second social distance from the user that is further than the first social distance in the social graph. A first graphical representation is provided in the network-based communication environment displayed to the first end-user and a second graphical representation different from the first graphical representation is provided in the network-based communication environment displayed to the second end-user. The difference between the first graphical representation and the second graphical representation is based on the difference between the first social distance and the second social distance.

Abstract: A method for multi-touch detection in a touch input device comprising a grid based sensor, the method comprises sampling outputs from a grid based sensor over a first stage of detection, selecting sensor lines based on the sampled outputs, scanning the selected sensor lines along one axis of the grid based sensor over a second stage of detection, and determining positions of user interaction based on outputs sampled in response to scanning the selected sensor lines during the second stage of detection.

Abstract: In non-limiting examples of the present disclosure, systems, methods and devices for determining a language of a text string are presented. A language detection model may be maintained. The language detection model may comprise identities and weights for initial and final consonants, identities and weights for prefixes and suffixes, and identities and weights for vowel sequences, where each identity is derived from a training corpus. The weights may correspond to a frequency of a text unit in the corpus. A text string may be received and a match score between the text string and the language of the language detection model may be determined. The match score may be based on initial and final consonant scores, prefix and suffix scores, and/or vowel sequence scores for each word in the text string. If the match score meets a threshold value a follow-up action associated with the language may be performed.

Abstract: The techniques disclosed herein provide a system that can identify two or more independently managed message threads and modify a user interface to enable a common user of the threads to view messages of both threads. For instance, if a user is participating in a first chat session that is displayed within a first message thread, the system can identify other threads that are related to the first thread. The related threads can be identified by a user input or by a query performed by a system. In some embodiments, the system can only search for threads that have at least one common user to the first thread. The system can provide this display of two related threads while linking the threads but not merging the messages of the threads. A data structure can be maintained to control the display of each thread.

Abstract: Some embodiments track internal actions of a development tool, and actions of a user of the tool. When the user faces a problem, information about the problem state is matched to a mitigation graph that was built from other user's sessions with the tool. Then documentation contextualized to the problem is synthesized, e.g., using machine learning models, and presented to the user. In addition to text and images, the documentation may include interaction points such as hyperlinks or buttons associated with code that mitigates the problem. For example, a mitigation sequence may proactively instruct the tool to perform certain actions such as an install or build or undo, or may automatically operate the tool user interface proactively in place of the user. Mitigation sequences may be sorted, and subjected to cost or outcome desirability prioritization. An administration tool provides mitigation graph management functionality. Personally identifiable information presence is constrained.

Abstract: Disclosed is a scalable, graph-based approach to detecting anomalous accesses to resources in a computer network. Access events are represented as edges between resource nodes and accessing nodes (e.g., corresponding to users) in a bipartite graph, from which vector representations of the nodes that reflect the connections can be computed by graph embedding. For an access event of interest, an anomaly score may be computed based on dissimilarities, in terms of their embedding distances, between the associated accessing node and other accessing nodes that have accessed the same resource, and/or between the associated resource node and other resource nodes that have been accessed by the same accessing node.

Abstract: Embodiments are provided for securing data access to machine learning training data at a plurality of distributed computing devices. Electronic content including original data that corresponds to a preferred data security level is divided into a plurality of microsegments. The plurality of microsegments is restrictively distributed to a plurality of computing devices which apply transcription labels to the plurality of microsegments. The labeled microsegments are reconstructed into training data which is then used to train a machine learning model while facilitating an improvement in data security of the original data included with the training data from the reconstructed microsegments.

Abstract: A camera system comprises a lamp configured to emit light, a lamp driver configured to energize the lamp, an image-sensor array configured to acquire image data, a lenslet array, and an image engine. The lenslet array comprises a plurality of lenslets arranged laterally over the image-sensor array and configured to focus the light, reflected from a subject, onto the image-sensor array. The image engine is configured to (a) receive the image data from the image-sensor array, (b) resolve the image data into a plurality of component images, and (c) return an enhanced image based on the plurality of component images, the enhanced image having enhanced spatial resolution relative to any of the component images.

Abstract: Computing resource management is improved during fast sorting using vector instructions. The process includes: determining a pivot value and a pivot position in a data set (e.g., by sampling with vectors and determining the sample median), determining whether moving data in the sampled portion may be avoided (e.g., if it is constant-valued or already sorted) and, leveraging that determination to possibly avoid unnecessary data movement, sorting the data set. Some examples further determine the microarchitecture version of the computing device performing the sorting and select an implementation of sorting instruction that is tuned for that microarchitecture version (e.g., based on the number of vector registers and motherboard cache configuration). Some examples leverage a soft 3-way quicksort by finding data elements adjacent to the pivot position that also have the pivot value and adding a partition boundary at the end of the set of same-valued data elements.

Abstract: A method, computer program product, and computing system for generating a plurality of acoustic relative transfer functions for a plurality of audio acquisition devices of an audio recording system deployed in an acoustic environment. The plurality of acoustic relative transfer functions may be encoded into a first embedding of acoustic relative transfer functions and at least a second embedding of acoustic relative transfer functions. Information may be extracted from at least the first embedding of acoustic relative transfer functions.

Abstract: A resource server system granting to users access to a resource based on the very fact that the users' computing systems can demonstrate that they heard an audio signal. Specifically, the resource server system detects receipt of a message from a client computing system, and interprets the message as representing that the client computing system heard an audio signal. In response, the resource server system grants a user of the client computing system access to the resource. This may be performed for multiple client computing systems that each demonstrate that they heard the audio signal. Thus, the principles described herein allow for the granting of access to resources to other computing systems within the audible proximity of a computing system that transmitted the audio signal.

Abstract: Examples are disclosed that relate to display assemblies and counterbalance mechanisms for rotatable displays. In one example, a counterbalance mechanism for a rotatable display comprises at least one crank arm rotatably connected to a ground. The counterbalance mechanism also comprises at least one spring comprising a first end connected to a spring end of the at least one crank arm, and a second end connected to the ground. At least one connecting link has a first portion connected to the at least one crank arm and a second portion connected to the rotatable display.

Abstract: A computing system for a plurality of classes of audio events is provided, including one or more processors configured to divide a run-time audio signal into a plurality of segments and process each segment of the run-time audio signal in a time domain to generate a normalized time domain representation of each segment. The processor is further configured to feed the normalized time domain representation of each segment to an input layer of a trained neural network. The processor is further configured to generate, by the neural network, a plurality of predicted classification scores and associated probabilities for each class of audio event contained in each segment of the run-time input audio signal. In post-processing, the processor is further configured to generate smoothed predicted classification scores, associated smoothed probabilities, and class window confidence values for each class for each of a plurality of candidate window sizes.

Abstract: Efficient tagging of content items using content embeddings are provided. In one technique, multiple content items are stored a content embedding for content item is stored. Entity names are also stored along with an entity name embedding for each entity name. For each content item, (1) multiple content embeddings that are associated with the content item are identified; (2) a subset of the entity names is identified; and (3) for each entity name in the subset, (i) an embedding of the entity name is identified, (ii) similarity measures are generated based on the entity name embedding and the multiple content embeddings, (iii), a distribution of the similarity measures is generated, (iv) feature values are generated based on the distribution, (v) the feature values are input into a machine-learned classifier, and (vi) based on output from the classifier, it is determined whether to associate the entity name with the content item.

Abstract: A column in a memory array includes one bit cell circuit in each row for storing information about the row. The bit cell circuits store data in a data node and a complement data node in a cross-coupled inverter circuit. Toggling the nodes in a cross-coupled inverter includes discharging a charged node in contention with a charge current provided in the cross-coupled inverter circuit. The column circuit includes a first voltage supply circuit to decouple the cross-coupled inverter circuits from a first voltage rail to cut off charging current in response to a column set signal in a column set operation. The cross-coupled inverters of the bit cell circuits in the column circuit are coupled to a first voltage supply line. The column set input controls the first voltage supply circuit to decouple the first voltage supply line from a first voltage rail.

Abstract: Systems and methods directed to returning personalized image-based search results are described. In examples, a query including an image may be received, and a personalized item embedding may be generated based on the image and user profile information associated with a user. Further, a plurality of candidate images may be obtained based on the personalized item embedding. The candidate images may then be ranked according to a predicted level of user engagement for a user, and then diversified to ensure visual diversity among the ranked images. A portion of the diversified images may then be returned in response to an image-based search.

Abstract: Methods and systems for transferring an application between a server environment and a local device are provided. Some examples include launching the application in the server environment. The application is associated with a URL. Some examples further include determining that the application should be executed on the local device, generating a transfer control option corresponding to switching the application from the server environment to the local device, receiving a selection of the transfer control option, and in response to receiving the selection, transmitting the URL from the server environment to the local device to enable execution of the application on the local device.

Abstract: Innovations in adaptive encoding of screen content based on motion type are described. For example, a video encoder system receives a current picture of a video sequence. The video encoder system determines a current motion type for the video sequence and, based at least in part on the current motion type, sets one or more encoding parameters. Then, the video encoder system encodes the current picture according to the encoding parameter(s). The innovations can be used in real-time encoding scenarios when encoding screen content for a screen sharing application, desktop conferencing application, or other application. In some cases, the innovations allow a video encoder system to adapt compression to different characteristics of screen content at different times within the same video sequence.

Abstract: A computing device includes a first portion and a second portion rotatable relative to each other about a hinge. A bumper is located between the hinge and the first portion. The bumper includes an impact resistant material. The impact resistant material absorbs some or all of the force from an impact between a hinge body and the first portion.

Abstract: Disclosed in some examples are methods, systems, devices, and machine-readable mediums which provide for sidebar communication threads forked from, or related to, a principal thread. Messages in the sidebar communication thread may include a history of the principal thread, including one or more messages from the principal thread, and may include a proposed principal thread message that is the subject of the sidebar thread discussion. The sidebar thread may also include sidebar thread messages that carries the conversation of the sidebar thread participants. Once a termination condition is reached for the sidebar thread, the sidebar thread terminates and either the proposed principal thread message (as potentially modified by participants of the sidebar thread) becomes an accepted principal thread message and it is posted to the principal thread as if it was sent by the sidebar initiator or no message is posted (e.g., the proposed principal thread message is rejected).

Abstract: Innovations in syntax and semantics of coded picture buffer removal delay (“CPBRD”) values potentially simplify splicing operations. For example, a video encoder sets a CPBRD value for a current picture that indicates an increment value relative to a nominal coded picture buffer removal time of a preceding picture in decoding order, regardless of whether the preceding picture has a buffering period SEI message. The encoder can signal the CPBRD value according to a single-value approach in which a flag indicates how to interpret the CPBRD value, according to a two-value approach in which another CPBRD value (having a different interpretation) is also signaled, or according to a two-value approach that uses a flag and a delta value. A corresponding video decoder receives and parses the CPBRD value for the current picture. A splicing tool can perform simple concatenation operations to splice bitstreams using the CPBRD value for the current picture.

Abstract: Innovations in syntax and semantics of coded picture buffer removal delay (“CPBRD”) values potentially simplify splicing operations. For example, a video encoder sets a CPBRD value for a current picture that indicates an increment value relative to a nominal coded picture buffer removal time of a preceding picture in decoding order, regardless of whether the preceding picture has a buffering period SEI message. The encoder can signal the CPBRD value according to a single-value approach in which a flag indicates how to interpret the CPBRD value, according to a two-value approach in which another CPBRD value (having a different interpretation) is also signaled, or according to a two-value approach that uses a flag and a delta value. A corresponding video decoder receives and parses the CPBRD value for the current picture. A splicing tool can perform simple concatenation operations to splice bitstreams using the CPBRD value for the current picture.

Abstract: An imaging system is configured to use an array of time-of-flight (ToF) pixels to determine depth information using the ToF imaging method and/or the stereo imaging method. A light emitting component emits light to illuminate a scene and a light detecting component detects reflected light via the array of ToF pixels. A ToF pixel is configured to determine phase shift data based on a phase shift between the emitted light and the reflected light, as well as intensity data based on an amplitude of the reflected light. Multiple ToF pixels are shared by a single micro-lens. This enables multiple offset images to be generated using the intensity data measured by each ToF pixel. Accordingly, via a configuration in which multiple ToF pixels share a single micro-lens, depth information can be determined using both the ToF imaging method and the stereo imaging method.

Abstract: The description relates to hinged devices, such as hinged computing devices. One example can include first and second portions that rotate around a hinge shaft that is fixedly secured to the first portion and rotationally secured to the second portion. The second portion defining a first contact surface spaced apart from a second contact surface. Multiple friction clips friction fit around the hinge shaft and rotating with the hinge shaft between the first contact surface and the second contact surface.

Abstract: Techniques for aligning images generated by two cameras are disclosed. This alignment is performed by computing a relative 3D orientation between the two cameras. A first gravity vector for a first camera and a second gravity vector for a second camera are determined. A first camera image is obtained from the first camera, and a second camera image is obtained from the second camera. A first alignment process is performed to partially align the first camera's orientation with the second camera's orientation. This process is performed by aligning the gravity vectors, thereby resulting in two degrees of freedom of the relative 3D orientation being eliminated. Visual correspondences between the two images are identified. A second alignment process is performed to fully align the orientations. This process is performed by using the identified visual correspondences to identify and eliminate a third degree of freedom of the relative 3D orientation.

Abstract: A deep learning model trained to learn to predict source code is tuned for a target source code generation task through reinforcement learning using a reward score that considers the quality of the source code predicted during the tuning process. The reward score is adjusted to consider code-quality factors and source code metrics. The code-quality factors account for the predicted source code having syntactic correctness, successful compilation, successful execution, successful invocation, readability, functional correctness, and coverage. The source code metrics generate a score based on how close the predicted source code is to a ground truth code.

Abstract: Video decoding innovations for multithreading implementations and graphics processor unit (“GPU”) implementations are described. For example, for multithreaded decoding, a decoder uses innovations in the areas of layered data structures, picture extent discovery, a picture command queue, and/or task scheduling for multithreading. Or, for a GPU implementation, a decoder uses innovations in the areas of inverse transforms, inverse quantization, fractional interpolation, intra prediction using waves, loop filtering using waves, memory usage and/or performance-adaptive loop filtering. Innovations are also described in the areas of error handling and recovery, determination of neighbor availability for operations such as context modeling and intra prediction, CABAC decoding, computation of collocated information for direct mode macroblocks in B slices, reduction of memory consumption, implementation of trick play modes, and picture dropping for quality adjustment.

Abstract: Embodiments of the present disclosure include techniques for compressing data using a tree encoded bit mask that may result in higher compression ratios. In one embodiment, an input vector having a plurality of values is received by a first plurality of switch circuits. Selection of the input values is controlled by sets of bits from the bit mask. The sets of bits specify locations of portions of the input vector where particular value of interest reside. The switch circuits output multiple values of the input vector, which include the particular value of interest. A second stage of switch circuits is controlled by logic circuit that detects values on the outputs of the first stage of switch circuits and outputs the values of interest. In some embodiments, the values of interest may be non-zero values of a sparse input vector, and the switch circuits may be multiplexers.

Abstract: A data processing system implements techniques for restricting which notifications and/or conversations are presented on a plurality of user devices associated with a user. Each user device is associated with a device category. The device categories are associated with time category information that associates each of a plurality of time periods with permitted device category information that indicates which categories of user devices associated with the user are permitted to present notifications that messages have been received for a particular time period and the categories of messages for which the notifications may be presented for that time period and/or which categories of conversations may be presented or hidden. The message category may be determined based on user input or by analyzing the message content with a machine learning model configured to predict the message category.