Abstract: A computer-implemented method includes accessing metadata that indicates how data sets that include a specified type of data are mapped among multiple data stores within a data management platform. The method includes auditing the data sets, based on the mapping, to determine where, among the data stores, instances of the specified type of data are stored. The method further takes an offline snapshot of the data stores that include the specified type of data to capture a current state of the instances of the specified type of data that were identified during the auditing and searches the offline snapshot for instances of the specified type of data that are marked for deletion. The method also deletes the marked instances of the specified type of data according to a data deletion policy that governs how the instances are to be deleted. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, an adaptive bitrate ladder (ABL) application limits streaming bitrates during live events. The ABL application receives a value from an edge server included in a CDN and determines a policy based on the value. The policy includes a maximum bitrate for streaming downloadables associated with a live event. The ABL application transmits the policy to the edge server. The edge server implements the policy to prohibit a client device from requesting a first segment of a first downloadable that is associated with the live event and has a bitrate that exceeds the maximum bitrate.

Abstract: A computer-implemented method for simulating web traffic to sandbox-test a new digital content platform service or feature. For example, implementations described herein identify and clone live production traffic from a first route including an existing digital content service. The implementation further forks the cloned production traffic along a second route to a new digital content service. By monitoring and correlating production responses from both the first and second routes, the implementations described herein can analyze and compare performance, accuracy, and correctness of the new digital content service to determine whether the new digital content service can handle live production traffic at scale. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented methods and systems can enable simultaneous interactions with a single media item via two or more display devices. For example, the methods and systems discussed herein pair two or more display devices under the same digital content system account. Once paired, the methods and systems discussed herein enable interactions with media items via the two or more display devices. For example, the methods and systems enable a viewer to watch a media item on a first display device while scrubbing to a different playback position within the media item on a second display device. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented methods and systems trigger actions by a client media player in response to determining that a live event has occurred during a live media broadcast. For example, the disclosed methods and systems determine that a live event has occurred during a live media broadcast and can generate instructions for various actions that the client media player should take in response to the live event. In one or more implementations, the disclosed methods and systems communicate these tailored instructions via the same dedicated communication channel that carries media content requests and responses. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a computer-implemented method comprises receiving a structured query identifying a plurality of source datasets, determining, based on a plurality of schemas associated with the plurality of source datasets, the plurality of source datasets associated with the structured query, loading a plurality of in-memory datasets, each in-memory dataset representing a source dataset in the plurality of source datasets, and executing the structured query on each of the plurality of in-memory datasets by causing portions of the structured query to be applied to each of the plurality of in-memory datasets to generate a query result.

Abstract: In various embodiments, a technique for reducing banding artifacts in decoded video data includes receiving a first set of reference samples, determining that the first set of reference samples meets a first criterion, selecting a first filter corresponding to the first criterion, applying the first filter to the first set of reference samples to generate a first set of filtered samples, performing at least one intra-prediction decoding operation on the first set of filtered samples to generate a first set of predicted samples, and generating a first portion of reconstructed video data based on the first set of predicted samples.

Abstract: The disclosed computer-implemented methods and systems include training a machine-learning model to accurately generate representations of similar scenes from long-form videos that have semantically different speech audio. For example, the methods and systems described herein generate machine-learning model training data including video clips and corresponding audio spectrograms. To augment this data, the methods and systems described herein further include dubbed audio spectrograms with the training data such that each video clips corresponds with a primary language audio spectrogram and a secondary language audio spectrogram. By applying a machine-learning model to this training data, the systems and methods described herein teach the machine-learning model to de-emphasize speech audio when generating audio visual representations corresponding to scenes from long-form video. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a data optimization application mitigates scoring inaccuracies in subjective quality experiments. In operation, the data optimization application generates a model that includes a first set of individual scores and a first set of parameters. The first set of parameters includes a first subjective score set and a first set of subjective factor sets. The data optimization application performs one or more optimization operations on the first set of parameters to generate a second set of parameters. The second set of parameters includes a second subjective score set and a second set of subjective factor sets, wherein the second subjective score set compensates for at least a first subjective factor set included in the second set of subjective factor sets.

Abstract: One embodiment sets forth a technique for automatically identifying clusters within a network to be tested for resilience. The technique includes the steps of determining that a first condition precedent for testing a first cluster within the network for resilience has been satisfied; determining that the first cluster should be tested for resilience based on a first property associated with the first cluster; in response to determining that the first condition precedent has been satisfied and determining that the first cluster should be tested for resilience, generating a cluster resilience test package for the first cluster; automatically causing the first cluster to be tested under a cluster resilience test in accordance with the cluster resilience test package. Another embodiment sets forth techniques for automatically carrying out the cluster resilience test in accordance with the cluster resilience test package.

Abstract: In various embodiments, a computer-implemented method comprises receiving, from a first endpoint device, a first event during a first time period, modifying an event log to include a record associated with the first event, causing a rollup queue to include a request to count a first count value associated with the first event, and generating, in a second time period subsequent to the first time period, a counter value associated with at least the first event based on the rollup queue and the event log.

Abstract: In various embodiments, an intent-based query processing application processes search queries. The intent-based query processing application computes lexical similarity scores between a search query and a set of entities. The intent-based query processing application computes entity relevance scores based on the lexical similarity scores and user engagement scores associated with both the search query and the set of entities. The intent-based query processing application computes a first category relevance score associated with both the search query and a first category based on the entity relevance scores. The intent-based query processing application determines an intent associated with the search query based on the first category relevance score. The intent-based query processing application generates a response to the search query based on the intent.

Abstract: The disclosed computer-implemented methods and systems optimize a rendering pipeline for rendering tasks associated with streamed video games. For example, the disclosed methods and systems amortize the fixed costs of these rendering tasks across multiple video game frames by combining a batch of frames into a single combined image. The disclosed systems and methods then perform rendering tasks such as frame generation and upscaling in connection with the single combined image. The disclosed system and methods further separate the combined image into its component frames prior to encoding and streaming those frames to one or more streaming clients. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments a training application trains convolutional neural networks (CNNs) to reduce reconstruction errors. The training application executes a first CNN on a source image having a first resolution to generate a downscaled image having a second resolution. The training application executes a second CNN on the downscaled image to generate a reconstructed image having the first resolution. The training application computes a reconstruction error based on the reconstructed image and the source image. The training application updates a first learnable parameter value included in the first CNN based on the reconstruction error to generate at least a partially trained downscaling CNN. The training application updates a second learnable parameter included in the second CNN based on the reconstruction error to generate at least a partially trained upscaling CNN.

Abstract: In various embodiments a computer-implemented method for transmitting frames of media content to a client device. The method includes generating a frame for a portion of media content, generating a header for the frame based on the portion of the media content, and generating one or more metadata blocks for the frame corresponding to a media encoding format, where the one or more metadata blocks include metadata incompatible with the header. The method further includes incorporating the header into a header portion of the frame, incorporating the portion of media content and the one or more metadata blocks into a payload portion of the frame, and transmitting the frame to a client device for playback.

Abstract: In various embodiments, a segment delivery application streams segments of downloadables to client devices. At a first point-in-time, the segment delivery application receives a request from a server for a segment of a downloadable. The segment delivery application determines that the segment is not available and that the segment is a next expected segment of the downloadable. At a second-point in time, the segment delivery application determines that the segment has become available. Upon determining that the segment has become available, the segment delivery application transmits to the server a response that includes the segment and corresponds to the request.

Abstract: The disclosed computer-implemented method includes accessing media frame generation input events produced as part of a multimedia application on a media server, selecting at least one media frame that is to be rendered according to the media frame generation input events, determining graphics processing capabilities of a client device on which the selected media frame is to be rendered, and generating a render command for the selected media frame based on the determined graphics processing capabilities of the client device. The render command includes contextual graphics information and graphics processing unit (GPU) pipeline information for use in rendering the selected media frame on the client device. The method also includes transmitting the generated render command to the client device to initiate rendering of the selected media frame using the contextual graphics information and the GPU pipeline information. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented method may include accessing a media item that includes multiple media item segments that are to be played back in a specific manner. The method may also include generating playgraphs for the media item, where the playgraphs define different playback paths between media item segments. The method may next include selecting a specific playgraph from the generated playgraphs, and then providing the selected playgraph to a playback device. Playback of the media item according to the selected playgraph may thereby provide a customized presentation of the media item. Various other methods, systems, and computer-readable media are also disclosed.