Abstract: Systems, methods, and computer-readable media are disclosed for using machine learning to generate automated custom media experiences. Example methods may include determining a first user identifier of a first user present in an environment, determining a first set of media preferences of the first user using media presented in association with the first user identifier, and determining a second user identifier of a second user present in the environment. Methods may include determining a second set of media preferences of the second user using media presented in association with the second user identifier, generating, using a first machine learning model, a first playlist of media using the first set of media preferences and the second set of media preferences for presentation to the first user and the second user, the first playlist comprising first media and second media, and causing audible presentation of the first media.

Abstract: Systems and techniques are generally described for detecting evasive terms in item listings. In some examples, a list of keywords associated with a first category of specified items may be received. In some examples, text data associated with a first item may be received. A first plurality of tokens representing a corrected version of the text data may be generated using spell correction techniques. In some cases, first flag data indicating that the text data constitutes evasive text may be generated based at least in part on the generation of the first plurality of tokens. The first flag data and a numerical representation of the first plurality of tokens may be input into a binary classifier trained to generate label data indicating whether text data associated with a given item is evasive or non-evasive. The binary classifier may generate label data indicating that the text data associated with the first item is evasive.

Abstract: Techniques for using suppression filters for presenting content of a network documents are described. In an example, a computer system receives, from a device associated with an account identifier, a request for an online network document. The computer system determines that a first category identifier of a first item category is associated with a suppression filter. The computer system determines that the account identifier is associated with an online item interaction having an interaction time associated with an item that belongs to the first item category. The computer system determines, based on a comparison of the interaction time and the suppression time, that a presentation of the first content in the online network document is prohibited. The computer system sends, to the device in response to the request, network document data that indicates second content about a third item that belongs to a second item category.

Abstract: Devices and techniques are generally described for generating per-topic review summarization. In some examples, a plurality of text descriptions of a first item may be received. In at least some examples, a list of categories may be determined. In some examples, at least one machine learned model may generate a first summary of the plurality of text descriptions. In some examples, the first summary may be relevant to a first category of the list of categories. In further examples, the at least one machine learned model may generate a second summary of the plurality of text descriptions. The second summary may be relevant to a second category of the list of categories. In various cases, first output data that includes the first summary and the second summary may be generated.

Abstract: Techniques are generally described for machine learning exampled-based annotation of image data. In some examples, a first machine learning model may receive a query image comprising a first depiction of an object-of-interest. In some examples, the first machine learning model may receive a target image representing a scene in which a second depiction of the object-of-interest is visually represented. In various examples, the first machine learning model may generate annotated output image data that identifies a location of the second depiction of the object-of-interest within the target image. In some examples, an object detection model may be trained based at least in part on the annotated output image data.

Abstract: A computer system and associated methods are disclosed for mitigating side-channel attacks using a shared cache. The computer system includes a host having a main memory and a shared cache. The host executes a virtual machine manager (VMM) that determines respective security keys for a plurality of co-located virtual machines (VMs). A cache controller for the shared cache includes a scrambling function that scrambles addresses of memory accesses performed by threads of the VMs according to the respective security keys. Different cache tiers may implement different scrambling functions optimized to the architecture of each cache tier. Security keys may be periodically updated to further reduce predictability of shared cache to memory address mappings.

Abstract: Systems and methods are provided for management of data access latency in a distributed storage system by dynamically replicating data objects experiencing a high request volume. When a storage node is unable to meet a data access latency target (e.g., due to one or more data objects hosted by the storage node being the subject of a large quantity of requests), the storage node may initiate a process by which a second storage node is employed to reduce the request processing load. The second storage node may cache a copy of the requested data object, and may service at least a portion of requests for the data object using the cached copy.

Abstract: Voice-controlled devices that include one or more speakers for outputting audio. In some instances, the device includes at least one speaker within a cylindrical housing, with the speaker aimed or pointed away from a microphone coupled to the housing. For instance, if the microphone resides at or near the top of the cylindrical housing, then the speaker may point downwards along the longitudinal axis of the housing and away from the microphone. By pointing the speaker away from the microphone, the microphone will receive less sound from the speaker than if the speaker were pointed toward the microphone). Because the voice-controlled device may perform speech recognition on audio signals generated by the microphone, less sound from the speaker represented in the audio signal may result in more accurate speech recognition, and/or a lesser need to perform acoustic echo cancelation (AEC) on the generated audio signals.

Abstract: Methods, systems, and computer-readable media for global column indexing in a graph database are disclosed. A plurality of data elements of a graph database are stored. The triples comprise identifiers, column names, and values. The column names are globally scoped in the graph database and are associated with data types. Indices corresponding to the column names are created. A particular one of the indices comprises one or more of the values associated with the corresponding column name. A query is performed on the graph database using one or more of the indices corresponding to one of more of the column names associated with the query.

Abstract: Techniques for training a visual neural network are described. In particular, the training of the visual neural network may be performed using one or more contrastive learning loss functions including one or more of a visual-to-visual contrastive learning function using the visual neural network on positive and negative video clips according to a first loss function, a secondary-to-secondary contrastive learning a secondary neural network on secondary positive and negative information generated from the positive and negative video clips, and a secondary-to-visual contrastive learning according to a third loss function using the visual neural network on positive and negative video clips and using the secondary neural network secondary positive and negative information generated from the positive and negative video clips.

Abstract: Disclosed are systems and methods to deliver location restricted content to devices that do not have a location component, such as a Global Positioning System component. A location aware device may send to a content delivery service, location information determined by a location component of the location aware device and receive, from the content delivery service, an encrypted access token. The location aware device may then send the access token to a non-location aware device that is on a same local area network as the location aware device. The non-location aware device may provide the access token to the content delivery service along with a request for location restricted content. The content delivery service may use the access token provided by the non-location aware device to verify that the non-location aware device is within a defined area for which the requested location restricted content is allowed for presentation.

Abstract: An apparatus includes a memory and a hardware processor. The hardware processor determines one or more of a weight of an item, a packaging type of the item, a packaging material of the item, a barcode of the item, a rigidity of the item, or a physical response of the item to being lifted and determines a visual appearance of the item and a shape or size of the item. The hardware processor also compares, using a machine learning model, the determined characteristics of the item to a manifest for the container. The manifest identifies a plurality of items in the container. The hardware processor determines, using the machine learning model, an identity of the item based on comparing the determined characteristics of the item to the manifest.

Abstract: Techniques for generating geographical units that can be used to generate delivery routes are described herein. Geospatial vector data and barrier geospatial vector data for a geographical area may be obtained. Seed points for one or more portions of the geographical area may be determined based at least in part on historical delivery volume for the geographical area. A plurality of polygons that represent the geographical area may be determined based at least in part on an algorithm that uses the seed points, the geospatial vector data, and the barrier geospatial vector data. Coordinates for a geographical unit of a plurality of geographical units that divide the geographical area may be determined based at least in part on the plurality of polygons and a polygon-to-polygon barrier aware drive time matrix that identifies a calculated cost for traveling from one polygon to another polygon using barriers identified in the barrier geospatial vector data.

Abstract: Systems and methods for optimizing a transportation plan for a transportation network are described. In an example, a computer system determines a forecasted demand for transporting items between a source node and a destination node of a transportation network and a set of constraints associated with a handling node of the transportation network. The computer system generates a transportation plan that comprises a network path including the handling node and a type of item handling at the handling node. In a first stage of optimization, an optimization model of the computer system selects the network path from network paths. In a second stage of optimization, the optimization model selects the type of item handling from types of item handling available at the handling node. The optimization model uses the network paths and the types of item handling as variables.

Abstract: Techniques for performing speech processing using multi-modal widget information are described. A system may receive input data corresponding to a user input. The system may also receive widget context data corresponding to one or more multi-modal widgets active at a device. The system may use the widget context data to perform natural language understanding (NLU) processing with respect to the user input, and for selecting a skill component for responding to the user input. The system may send a widget identifier to the skill component when invoking the skill to respond to the user input.

Abstract: The present disclosure generally relates to a system of a delivery device for combining sensor data from various types of sensors to generate a map that enables the delivery device to navigate from a first location to a second location to deliver an item to the second location. The system obtains data from RGB, LIDAR, and depth sensors and combines this sensor data according to various algorithms to detect objects in an environment of the delivery device, generate point cloud and pose information associated with the detected objects, and generates object boundary data for the detected objects. The system further identifies object states for the detected object and generates the map for the environment based on the detected object, the generated object proposal data, the labeled point cloud data, and the object states. The generated map may be provided to other systems to navigate the delivery device.

Abstract: Data is encoded to be incrementally authenticable. A plaintext is used to generate a ciphertext that comprises a plurality of authentication tags. Proper subsets of the authentication tags are usable to authenticate respective portions of plaintexts obtained from the ciphertext. Portions of the plaintext can be obtained and authenticated without decrypting the complete ciphertext.

Abstract: Techniques for change data capture (CDC) log augmentation are described. In some examples, a user configures CDC log augmentation by indicating which data should be included in a CDC log, and the database, when generating a CDC log associated with this configuration, can obtain the associated data and augment the CDC log by inserting this data into it. The augmented data can include one or more fields from a record in a separate database table, where the record can be identified based on the changed record represented by the CDC log.