Abstract: A system and method for method for providing access to a resource includes receiving, from a user device of a user, a request to access the resource and an access token, the access token identifying an access role associated with the user and the request identifying a task to be performed on the resource; retrieving the access role from the access token; examining a role-to-permissions map to identify one or more permission available for the access role; determining, based on the available permissions for the access role that the user is authorized to perform the task on the resource; determining, based on the access role and the available permissions that the user has permission to access the resource managed by an object model; and upon determining that the user is authorized to perform the task on the resource and the user has permission to access the resource managed by the object model, permitting the user device of the user to access the resource to perform the task.

Abstract: A method for classifying a tenant as being associated with one of a plurality of customer categories includes receiving a request to classify a tenant as being associated with one of a plurality of customer categories, dynamically retrieving from one or more data source systems, a plurality of data signals associated with the tenant, one or more of the plurality of data signals being stored as individual properties for the tenant in the one or more data source systems, dynamically identifying, in real-time, the tenant as being associated with one of the plurality of customer categories based at least on the plurality of data signals, and storing the identified one of the plurality of customer categories as a classification data signal for the tenant.

Abstract: A transformer is described herein for using transformer-based technology to process data items (e.g., image items). The transformer increases the efficiency of the transformer-based technology by using a modified attention component. In operation, the modified attention component accepts embedding vectors that represent a plurality of item tokens, together with a classification token. A first stage of the modified attention component generates original attention information based on the embedding vectors. A second stage generates score information based on a portion of the original attention information that pertains to the classification token. A third stage produces modified attention information by removing attention values from the original attention information, as guided by a sampling operation that is performed on the score information. The second and third stages do not rely on machine-trained values, which expedites the deployment of these functions in existing transformers.

Abstract: One example provides a computing device including a first portion including a first display, a second portion including a second display and a camera, the second portion connected to the first portion via a hinge, a hinge angle sensing mechanism including one or more sensors, a logic device, and a storage device holding instructions executable by the logic device to execute a camera application and to receive sensor data from the one or more sensors, based at least in part on the sensor data received from the one or more sensors, determine a device pose, output the camera application to the first display when the device pose is indicative of the camera being world-facing, and output the camera application to the second display when the device pose is indicative of the camera being user-facing.

Abstract: A method for training a language model comprises (a) receiving vectorized training data as input to a multitask pretraining problem; (b) generating modified vectorized training data based on the vectorized training data, according to an upstream data embedding; (c) emitting pretraining output based on the modified vectorized training data, according to a downstream data embedding equivalent to the upstream data embedding; and (d) adjusting the upstream data embedding and the downstream data embedding by computing, based on the pretraining output, a gradient of the upstream data embedding disentangled from a gradient of the downstream data embedding, thereby advancing the multitask pretraining problem toward a pretrained state.

Abstract: A computing device, including memory, an accelerator device, and a processor. The processor may generate a plurality of data packs that each indicate an update to a variable of one or more variables of a combinatorial cost function. The processor may transmit the plurality of data packs to the accelerator device. The accelerator device may, for each data pack, retrieve a variable value of the variable indicated by the data pack and generate an updated variable value. The accelerator device may generate an updated cost function value based on the updated variable value. The accelerator device may be further configured to determine a transition probability using a Monte Carlo algorithm and may store the updated variable value and the updated cost function value with the transition probability. The accelerator device may output a final updated cost function value to the processor.

Abstract: In a numerical simulation, input data expressed in at least a first domain is received. The input data is transformed to generate frequency modes of the input data in frequency domain. The transformed data is down-sampled to retain a subset of the frequency modes in the frequency domain. The down-sampled data is successively processed with one or more stages of a neural network to generate a down-sampled output in the frequency domain. The processing includes applying, in each stage of the one or more stages, a non-linear transformation to the subset of the frequency modes. The down-sampled output is then up-sampled to generate an up-sampled output corresponding to the frequency modes in the frequency domain, and the up-sampled output is transformed from the frequency domain to the at least the first domain to generate a result of the numerical simulation.

Abstract: Embeddable extensions of the functionality of email and calendaring applications are described herein. An originating email and calendaring application formats meeting items to include extensible content, e.g., within the body field of the meeting item. Recipient email and calendaring applications receive the meeting items and, depending on whether they are configured to provide an extended functionality, will either deploy the extensible content item according to the extended functionality or present the extensible content item according to a standardized functionality. The body field includes the extensible content item according to the standardized functionality that is interpretable by configured clients to provide the extended functionality.

Abstract: Conversational understanding systems allow users to conversationally interface with a computing device. In examples, a query may be received that includes a request for execution of a task. A data exchange task definition may be accessed. The data exchange task definition assists a conversational understanding system in managing task state tracking for information needed for task execution. Using the data exchange task definition, a per-turn policy for interacting with the user computing device is generated based on the state of a dialogue with a computing device and an evaluation of a process flow chart provided by a task owner resource. The task owner resource may be independent from the conversational understanding system. A response to the query may be generated and output based on the per-turn policy. In examples, the per-turn policy is used to generate one or more responses during a dialogue with a user via a computing device.

Abstract: Techniques of memory tiering in computing devices are disclosed herein. One example technique includes retrieving, from a first tier in a first memory, data from a data portion and metadata from a metadata portion of the first tier upon receiving a request to read data corresponding to a system memory section. The method can then include analyzing the data location information to determine whether the first tier currently contains data corresponding to the system memory section in the received request. In response to determining that the first tier currently contains data corresponding to the system memory section in the received request, transmitting the retrieved data from the data portion of the first memory to the processor in response to the received request. Otherwise, the method can include identifying a memory location in the first or second memory that contains data corresponding to the system memory section and retrieving the data from the identified memory location.

Abstract: In a numerical simulation, input data expressed in at least a first domain is received. The input data is decomposed into at least i) low-pass filtered data that captures a low-pass filtered version of the input data in the at least the first domain and ii) high-pass filtered data that captures a high-pass filtered version of the input data in the at least the first domain. The low-pass filtered data is transformed to frequency domain, and weights are applied to the low-pass filtered data in the frequency domain to generate weighted low-pass filtered data in the frequency domain. The weighted low-pass filtered data is transformed from the frequency domain to the at least the first domain, and output data for the numerical simulation is composed based on at least the weighted low-pass filtered data in the at least the first domain.

Abstract: A method of fabricating semiconductor-superconductor nanowires, comprising: forming a first mask amorphous mask having first openings over trenches in a substrate; forming a monocrystalline conducting material in the first openings by selective area growth, thus forming gates for the nanowires in the trenches pf the substrate; forming a second mask over the substrate and gates, the second mask also being amorphous and having a pattern of second openings; forming an insulating crystalline buffer in the second openings; forming a crystalline semiconductor material on the buffer in the second openings by selective area growth in order to form the cores of the nanowires, wherein the gates intersect with the cores in the plane of the substrate; and forming the coating of superconductor material over at least part of each of the cores.

Abstract: Examples are disclosed herein relating to signal processing in a time of flight (ToF) system. One example provides, a method comprising emitting, via a light source, amplitude-modulated light toward an object, acquiring, via an image sensor comprising a plurality of pixels, a plurality of image frames capturing light emitted from the light source that is reflected by the object, wherein the plurality of image frames are acquired at two or more different frequencies of the amplitude-modulated light and collectively form a multifrequency frame, and for each pixel of the multifrequency frame, determining a brightness level, applying an adaptive denoising process by setting a kernel size based on the brightness level, and performing a phase unwrapping process to determine a depth value for the pixel.

Abstract: One aspect provides semiconductor-superconductor hybrid device comprises a substrate, a first semiconductor component arranged on the substrate, a superconductor component arranged to be capable of energy level hybridisation with the first semiconductor component, and a second semiconductor component arranged as a gate electrode for gating the first semiconductor component. Another aspect provides a semiconductor-superconductor hybrid device, comprising: a substrate; a semiconductor component arranged on the substrate; a gate electrode for gating the semiconductor component; and a superconductor component capable of undergoing energy level hybridisation with the semiconductor component; wherein the gate electrode is arranged in a channel in the substrate. Also provided are methods of fabricating the semiconductor-superconductor hybrid devices.

Abstract: A computing device including a processor configured to simulate a quantum device at least in part by receiving a single-particle Hamiltonian matrix that describes an initial Hamiltonian operator. The initial Hamiltonian operator may model a plurality of parts of a quantum device. Simulating the quantum device may further include estimating a reduced density matrix associated with a first part, estimating a plurality of eigenvectors and eigenvalues of the reduced density matrix, and generating a transformed Hamiltonian matrix. Generating the transformed Hamiltonian matrix may include transforming the single-particle Hamiltonian matrix into a natural-orbital basis of the first part such that the transformed Hamiltonian matrix has a reduced dimensionality. The natural-orbital basis may be spanned by a subset of the eigenvectors of the reduced density matrix.

Abstract: A scheme for modifying an image is disclosed, which includes receiving a source image having a first image configuration; determining a second image configuration for a target image; providing the received source image to an AI engine trained to identify, based on a set of rules related to visual features, candidate regions from the source image; generating proposal images based on the candidate regions, respectively; determining, based on prior aesthetical evaluation data, an aesthetical value of each regional proposal image; and selecting, based on the determined aesthetical value of each regional proposal image, one of the regional proposal images as the target image; extracting, from the AI engine, the target image; and causing the target image to be displayed via a display of a user device.

Abstract: A system to detect an abnormal classic authorizations, such as in a classic authorization system of a resource access management system, and take action is described. The system determines an anomaly score in from a model applied to a classic assignment event. An indicator score is determined from the classic assignment event applied to domain-based rules. The security action is taken based on a combination of the anomaly score and the indicator score.

Abstract: To meet the stringent 5G radio access network (RAN) service requirements, layers one and two need to be processed in essentially real time. Thus, prompt anomaly detection is important to prevent negative impacts on customer experience, which is critical for mobile networks to meet the stringent service requirements. However, monitoring networks for anomalies is difficult due at least to (1) the resource constrained edge deployments in which the vRAN resides, (2) the variety of anomaly types and fault locations making anomalies difficult to detect, and (3) the low frequency of anomalies leading to unbalanced data sets for training, among others. The present application addresses these issues by decoupling anomaly detection at the infrastructure layer (servers, NICs, switches, etc.) from anomaly detection at the VNF layer (L1, high-DU, CU). This enables different techniques for identifying anomalies and for reducing the monitoring overhead that is tailored to each layer.

Abstract: A data processing system implements techniques for provisioning and operating databases including receiving a request to provision a new database that includes multiple replicas. The data processing system provisions the new database including the plurality of replicas. The plurality of replicas includes a primary replica, one or more secondary replicas, and one or more delta replicas. The primary replica is configured to handle read and write operations. The one or more secondary replicas include a copy of data and transactional log information of the primary replica and are unable to handle write operations. The one or more delta replicas are a copy of the transactional log information of the primary replica and are unable to handle read operations or write operations.

Abstract: Apparatus and methods for training a neural network accelerator using quantized precision data formats are disclosed, and in particular for storing activation values from a neural network in a compressed format having lossy or non-uniform mantissas for use during forward and backward propagation training of the neural network. In certain examples of the disclosed technology, a computing system includes processors, memory, and a compressor in communication with the memory. The computing system is configured to perform forward propagation for a layer of a neural network to produced first activation values in a first block floating-point format. In some examples, activation values generated by forward propagation are converted by the compressor to a second block floating-point format having a non-uniform and/or lossy mantissa. The compressed activation values are stored in the memory, where they can be retrieved for use during back propagation.

Abstract: A method for generating a comment capsule for a content segment includes receiving the comment related to the content segment, receiving data associated with the comment, the data including at least one of an identifier for the content segment to which the comment relates, an identifier for a user generating the comment, and a time at which the comment was generated, generating display data for displaying the comment in a comment capsule, and transmitting the display data for displaying the comment capsule in a user interface screen associated with the content segment. The comment capsule is displayed on the user interface screen inline with the content segment, and an appearance of the comment capsule is dynamically changeable based on at least one of a number of comments received for the content segment, a size of the content segment, and a read status of the comment.

Abstract: One example provides a computing device comprising a logic machine and a storage machine holding instructions executable by the logic machine to implement a depth image processing pipeline comprising a neural network, the neural network comprising an edge detecting layer. The neural network is configured to receive input of an active brightness image and receive input of one or more of real data or imaginary data of a complex depth image, the complex depth image corresponding to the active brightness image. The neural network is further configured to, at the edge detecting layer, apply one or more convolutional processes to the active brightness image to identify one or more edge pixels in the active brightness image, and at a second layer, denoise one or more of the real data or the imaginary data of the complex depth image based on the one or more edge pixels identified.

Abstract: A data processing system implements receiving, from a first user, a first request to submit a first template for an electronic document to a template publication service, the request includes first policy information indicating who may access the template and is available only to users who are associated with a first organization; storing the first template and the first policy information in a template data store; receiving, from a second user, a second request for available templates from the template data store; determining that the second user has access to the first template based on the policy information and the information identifying the second user indicating that the second user is associated with the first organization; including the first template in a plurality of first templates determined to be available to the second user; and causing the plurality of first templates available to the second user to be displayed.

Abstract: A system to detect abnormal cross authorizations and take action is described. The system determines whether cross authorization event applied to a first trained anomaly detection model and activity post cross authorization event applied to a second trained anomaly detection model is suspicious. An indicator score is determined from rule-based security indications applied to the cross authorization. A security action is taken based on application of the indicator score applied to a threshold.

Abstract: A computing device for organizing a collaborative workspace for a user implements providing, using a display of the computing device, an interactive user interface associated with the collaborative workspace in which the user may work on a shared task in a coordinated fashion with other users; identifying, based on one or more factors, one or more activities specific to the user from within the collaborative workspace; assigning the identified one or more activities to a catch-up list associated with the user; displaying, within the interactive user interface, the catch-up list; detecting that at least one of the one or more activities has been addressed from within the collaborative workspace or the catch-up list; and updating, based on the detection, the collaborative workspace and the catch-up list.

Abstract: A server computing device generates training data based upon an identifier for a device, a timestamp, and a label received from a developer computing device. The server computing device trains a computer-implemented machine learning (ML) model based upon the training data. The server computing device also generates client configuration data for the ML model that specifies transformations that are to be applied to values in order to generate input values for the ML model. The server computing device deploys ML assets to client computing devices, the ML assets comprising the ML model and the client configuration data. The client computing devices execute the ML model using input values derived via transformations of (local) values produced by the client computing devices and transmit telemetry data to the server computing device. The server computing device updates the ML assets based upon the telemetry data.

Abstract: Systems and methods are provided for determining faces and bodies of people in an image by adaptively scaling images and by iteratively using a deep neural network for inferencing. A camera captures an image including faces and bodies of people. A face/body determiner determines faces and bodies of people appearing in the image by resizing the image into a predetermined pixel dimension as input to the deep neural network. A region cropper determines a crop region associated with a low level of confidence in detecting faces and bodies that are too small to determine with an acceptable level of confidence. The region cropper resizes the crop region into the predetermined pixel dimension as input to the deep neural network. The face and body determiner determines other faces and bodies appearing in the resized crop region. An aggregator aggregates locations of the determined faces and bodies in the image.

Abstract: The discussion relates to disaggregated computing. One example can monitor multiple two-phase liquid immersion tanks. Individual two-phase liquid immersion tanks can contain multiple components of a single type of component type. The example can receive requests for virtual machines and allocate sets of components from individual two-phase liquid immersion tanks to work together to support the virtual machines requests.

Abstract: Aspects of the present disclosure relate to adjusting an illumination of a user depicted in one or more images when using a video conferencing application. In one example, one or more images depicting the user may be received from an image sensor. Further, an illumination of the user depicted in the one or more images may be determined to be unsatisfactory. For example, the user's face may be too bright or too dim. Accordingly, content displayed at a display device may identified and then modified. The modified content may then be rendered to a display device thereby changing the illumination of the user depicted in subsequent images. In examples, the modified content may include a graphical element, such as a ring of a specific color at least partially surrounding content rendered to and displayed at the display device.

Abstract: A computing system including a processor configured to receive training data including, for each of a plurality of training timesteps, training forecast states associated with respective training-phase agents included in a training supply chain graph. The processor may train a reinforcement learning simulation of the training supply chain graph using the training data via policy gradient reinforcement learning. At each training timestep, the training forecast states may be shared between simulations of the training-phase agents during training. The processor may receive runtime forecast states associated with respective runtime agents included in a runtime supply chain graph. For a runtime agent, at the trained reinforcement learning simulation, the processor may generate a respective runtime action output associated with a corresponding runtime forecast state of the runtime agent based at least in part on the runtime forecast states. The processor may output the runtime action output.

Abstract: Aspects of the present disclosure relate to techniques for minimizing the effects of RowHammer and induced charge leakage. In examples, systems and methods for preventing access pattern attacks in random-access memory (RAM) are provided. In aspects, a data request associated with a page table may be determined to be a potential security risk and such potential security risk may be mitigated by randomly selecting a memory region from a subset of memory regions, copying data stored in a memory region associated with a page table entry in the page table to the second memory region, disassociating the second memory region from the subset of memory regions and associating the memory region associated with the page table to the second memory region, and updating the page table entry in the page table to refer to the second memory region.

Abstract: A data processing system implements obtaining first textual content in a first language from a first client device; determining that the first language is supported by a first machine learning model; obtaining a guard list of prohibited terms associated with the first language; determining that the textual content does not include one or more prohibited terms associated based on the guard list; providing the first textual content as an input to the first machine learning model responsive to the textual content not including the one or more prohibited terms; analyzing the first textual content with the first machine learning model to obtain a first content recommendation; obtaining a first content recommendation policy that identifies content associated with the first language that may not be provided as a content recommendation; determining that the first content recommendation is not prohibited; and providing the first content recommendation to the first client device.

Abstract: A system hosts a file system representing a replica of file systems hosted on a plurality of nodes, each node storing data on a first storage medium. The system stores data on a second storage medium that is slower than the first storage medium. The system stores on the second storage medium a file received from one of the nodes when the file remains un-accessed by the nodes for a predetermined period of time. The first storage medium on the plurality of nodes stores a pointer to the file and not the data content of the file stored on the second storage medium. The system provides the file from the second storage medium to the first storage medium on one of the nodes while providing the plurality of nodes read and write access to the file stored on the second storage medium using the pointer on the first storage medium.

Abstract: A method comprising: storing, in a database, a list of geographical regions in which the party consents to their data being stored; receiving a request to store data of the party, wherein at least one of the data has a region-restriction that restricts the regions in which that data can be stored; determining, based on a respective region-restriction of a respective one of the data, in which of the list of regions the respective data can be stored; storing the data, wherein each respective data is stored in at least one respective storage centre associated with one of the regions according to the determination; storing, in the database, a list comprising a pointer to each respective stored data, wherein the pointer identifies the respective storage centre; receiving a request to retrieve a respective stored data; and using the pointer to route the request to the respective storage centre.

Abstract: Techniques performed by a data processing system for a machine learning driven teleprompter include displaying a teleprompter transcript associated with a presentation on a display of a computing device associated with a presenter; receiving audio content of the presentation including speech of the presenter in which the presenter is reading the teleprompter transcript; analyzing the audio content of the presentation using a first machine learning model to obtain a real-time textual translation of the audio content, the first machine learning model being a natural language processing model trained to receive audio content including speech and to translate the audio content into a textual representation of the speech; analyzing the real-time textual representation and the teleprompter transcript with a second machine learning model to obtain transcript position information; and automatically scrolling the teleprompter transcript on the display of the computing device based on the transcript position informatio

Abstract: This document generally relates to techniques for testing or training data augmentation. One example includes a method or technique that can include accessing a repository of private data items. The repository can provide a distribution of the private data items that is representative of a designated real-world scenario for a machine learning model. The method or technique can also include assigning classifications to the private data items in the repository. The method or technique can also include augmenting a testing or training set for the machine learning model based at least on the classifications of the private data items to obtain an augmented testing or training set that is relatively more representative of the distribution of classifications in the repository.

Abstract: A semiconductor-superconductor hybrid device comprises a semiconductor component and a superconductor component arranged over the semiconductor component. The superconductor component comprises a continuous portion of a superconductor material and a discontinuous portion of a non-ferromagnetic metal. The discontinuous portion is configured to increase the critical field of the superconductor component. It has been found that providing a superconductor component with a discontinuous portion of non-ferromagnetic metal may increase the critical field of the superconductor component, allowing the device to be operated in a stronger magnetic field. Further aspects provide a method of fabricating the device, and the use of a non-ferromagnetic metal to increase the critical field of a superconductor component of a semiconductor-superconductor hybrid device.

Abstract: An enterprise system server, a computer-readable storage medium, and a method for targeted training of inductive multi-organization recommendation models for enterprise applications are described herein. The method includes receiving enterprise application data from remote organization computing systems executing the enterprise application, training per-organization recommendation models for a subset of the organizations, and validating each per-organization recommendation model on enterprise application data corresponding to one or more other organizations.

Abstract: Aspects of the present disclosure relate to grounded multimodal agent interactions, where a natural language user input is processed using a multimodal machine learning model to generate model output. The model output may then be processed to affect the behavior of an application, for example to enable a user to control the application and/or to facilitate user interactions with a conversational agent, among other examples. In some instances, at least a part of the model output may be executed or parsed, for example to generate three dimensional objects within a coding application. Thus, use of a multimodal machine learning model according to aspects described herein may enable the use of user-provided natural language input to affect the behavior of an application accordingly.

Abstract: Aspects of the present disclosure relate to grounded multimodal agent interactions, where a user input is processed using a multimodal machine learning model to generate model output. The model output may then be processed to affect the behavior of an application, for example to enable a user to control the application and/or to facilitate user interactions with a conversational agent, among other examples. In some instances, at least a part of the model output may be executed or parsed, for example to call an application programming interface or function of the application. Thus, use of a multimodal machine learning model according to aspects described herein may enable the use of user-provided natural language input to affect the behavior of an application accordingly.

Abstract: Examples are disclosed that relate to head-mounted display (HMD) devices and methods for detecting light reflected by a user's eye. In one example, an HMD device comprises a frame, a transparent cover substrate supported by the frame, and a display substrate supported by the frame. An eye-tracking light source is affixed to the frame. The eye-tracking light source is configured to emit eye-tracking light. At least one input optical element is configured to receive the eye-tracking light. A plurality of transparent delivery waveguides are integrated with the transparent cover substrate. Each of the transparent delivery waveguides comprises an output optical element configured to output the eye-tracking light towards a user's eye. In addition, each of the transparent delivery waveguides comprises a curved portion. The HMD device further comprises a camera configured to detect the eye-tracking light reflected by the user's eye.

Abstract: Systems and methods for determining granular authorization of an authorization request for a user operation are provided. In particular, a computing device may receive the authorization request from a user requesting to perform the user operation on a resource, authenticate the user based on identity information included in the authorization request, evaluate the authorization request to identify one or more authorization claims for performing the user operation on the resource, determine one or more claim providers for generating the one or more authorization claims, and fetch the one or more authorization claims from the one or more claim providers. The computing device may further evaluate one or more authorization policies for determining whether the user is authorized to perform the user operation on the resource and determine permissions granted for the authorization request based on the evaluating the one or more authorization policies and the one or more authorization claims.

Abstract: Techniques are described for low-latency real-time streaming of media content. For example, streaming media content can be received from a media source, where the streaming media content comprises audio and/or video content. An audio/video stream can be streamed to one or more streaming clients. The audio/video stream is streamed as a sequence of encoded audio and/or video frames, which are independent encoded audio and/or video frames that are not grouped into chunks for streaming. The sequence of encoded audio and/or video frames is streamed to the one or more streaming clients as a one-way stream and without receiving any requests from the one or more streaming clients for subsequent frames or chunks.

Abstract: Aspects of the present disclosure relate to grounded multimodal agent interactions, where a user input is processed using a multimodal machine learning model to generate model output. The model output may then be processed to affect the behavior of an application, for example to enable a user to control the application and/or to facilitate user interactions with a conversational agent, among other examples. In some instances, at least a part of the model output may be executed or parsed, for example to call an application programming interface or function of the application. Thus, use of a multimodal machine learning model according to aspects described herein may enable the use of user-provided natural language input to affect the behavior of an application accordingly.

Abstract: In examples, a developer may define a set of computer-controlled agent attributes, which may be processed by a generative multimodal machine learning model in conjunction with background information associated with a virtual environment (e.g., “lore”) and other agent information to generate multimodal model output with which to control the behavior of the computer-controlled agent. Thus, a player may interact with the computer-controlled agent, such that user input from the player is processed using the ML model to generate model output to affect the behavior of the computer-controlled agent, thereby enabling the user and the computer-controlled agent to interact. As compared to manual dialogue authoring, use of agent information to define the behavior of a computer-controlled agent may result in reduced effort on the part of a creator while also offering increased depth and variety for computer-controlled agents of a virtual environment.

Abstract: Technology is described herein for searching an index, including operations of: obtaining a source data item; generating a source context-supplemented vector based on the source data item; and searching the index to find one or more target context-supplemented vectors that are determined to match the source context-supplemented vector. Each context-supplemented vector, which is associated with a particular data item, is made up of two parts: a language-agnostic vector and a context vector. The language-agnostic vector expresses the meaning of the particular data item in a manner that is independent of a natural language that is used to express the particular data item, while the context vector expresses a context associated with the formation of the particular data item. More generally, the technology’s use of context vectors allows it to perform index search operations in a more efficient manner, compared to a search engine that does not use context vectors.

Abstract: Aspects of the present disclosure relate to a machine learning-based dialogue authoring environment. In examples, a developer or creator of a virtual environment may use a generative multimodal machine learning (ML) model to create or otherwise update aspects of a dialogue tree for one or more computer-controlled agents and/or players of the virtual environment. For example, the developer may provide an indication of context associated with the dialogue for use by the ML model, such that the ML model may generate a set of candidate interactions accordingly. The developer may select a subset of the candidate interactions for inclusion in the dialogue tree, which may then be used to generate associated nodes within the tree accordingly. Thus, nodes in the dialogue tree may be iteratively defined based on model output of the ML model, thereby assisting the developer with dialogue authoring for the virtual environment.

Abstract: Examples of the present disclosure describe systems and methods for establishing a persistent connection between file attachments uploaded to cloud storage and associated electronic communications. An electronic communication including an attached file may be stored and presented in a communication application user interface that includes interactive functions, such as edit and save functions. Upon selection of an interactive function, a copy of the file may be uploaded to a cloud storage location and, in some examples, opened in a productivity application. A link to the file at the cloud storage location may be stored as a property of the file that is stored with the electronic communication. Resultantly, a connection may be maintained between the file at the cloud storage location and the electronic communication. Based on this connection, the copy of the file at the cloud storage location may be opened upon a subsequent selection of an interactive function.

Abstract: Technologies are described for categorization of time blocks in calendars. A time block to be categorized on a user's calendar may be determined based on a direct user input or inference from user history or behavior. Next, a category to be assigned, a beginning, and an end for the time block may be determined. The time block may be the same on all work days, different on each day, etc. One or more exceptions to be processed during the categorized time block may also be determined through direct user input or inference. Then, one or more rules associated with the assigned category restricting scheduling of events on the user's calendar during the categorized time block may be implemented subject to the one or more exceptions.

Abstract: This document relates to training and employing of quality estimation models to estimate the quality of different signal characteristics. One example includes a method or technique that can be performed on a computing device. The method or technique can include obtaining training signals exhibiting diverse impairments introduced when the training signals are captured or diverse artifacts introduced by different processing characteristics of a plurality of data enhancement models. The method or technique can also include obtaining quality labels for different signal characteristics of the training signals. The method or technique can also include training at least two different quality estimation models to estimate quality of at least two different signal characteristics based at least on the training signals and the quality labels.