Abstract: An online concierge system trains a classification model as a domain adversarial neural network from training data labeled with source classes from a source domain that do not overlap with target classes from a target domain output by the classification model. The online concierge system maps one or more source classes to a target class. The classification model extracts features from an image, classifies whether an image is from the source domain or the target domain, and predicts a target class for an image from the extracted features. The classification model includes a gradient reversal layer between feature extraction layers and the domain classifier that is used during training, so the feature extraction layers extract domain invariant features from an image.

Abstract: An online concierge system trains a classification model as a domain adversarial neural network from training data labeled with source classes from a source domain that do not overlap with target classes from a target domain output by the classification model. The online concierge system maps one or more source classes to a target class. The classification model extracts features from an image, classifies whether an image is from the source domain or the target domain, and predicts a target class for an image from the extracted features. The classification model includes a gradient reversal layer between feature extraction layers and the domain classifier that is used during training, so the feature extraction layers extract domain invariant features from an image.

Abstract: Certain aspects of the present disclosure are directed to techniques and apparatus for media coordination. One example method generally includes receiving, from a first user device, a command to play a voice message and media content at one or more second user devices, and determining whether one or more services accessible by the one or more second user devices can be used to play the media content. The method may also include coordinating the voice message and the media content to be played by the one or more second user devices based on the determination.

Abstract: Certain aspects of the present disclosure are directed to techniques and apparatus for media coordination. One example method generally includes receiving, from a first user device, a command to play a voice message and media content at one or more second user devices, and determining whether one or more services accessible by the one or more second user devices can be used to play the media content. The method may also include coordinating the voice message and the media content to be played by the one or more second user devices based on the determination.

Abstract: Embodiments are disclosed for caching multimedia content. An example in-vehicle computing system that adaptively caches multimedia data, the in-vehicle system including a memory, a position sensor, a receiver, and a processor. The position sensor provides a position signal indicative of a present location of the system. The receiver wirelessly receives multimedia data from a remote multimedia data source. The processor stores the received multimedia data in the memory, transfers multimedia data in the memory to a queue for playback when an amount of multimedia data saved in the memory is at least a first threshold value, and adjusts a size of the memory when a wireless communication channel is not available between the system and a remote multimedia data source.

Abstract: Described are techniques for processing a request from a client for performing an operation that is received by a business logic provider. First processing is performed by the business logic provider to process the request that includes communicating with other providers for information in connection with the request, performing validation processing using the information obtained from the other providers to determine whether one or more restrictions are violated, each of said one or more restrictions identifying a required state of one or more objects for performing an operation of the request, determining whether any restrictions are violated, and performing one or more actions responsive to determining that any restrictions are violated.

Abstract: Described are techniques for processing a request from a client. Validation processing is performed to determine whether any restrictions are violated. With respect to two associated objects for a logical device and its snapshot, a business logical provider may perform runtime checks regarding the two associated objects of the same or different frameworks to ensure a valid state exists with respect to the requested operation on the two associated objects. For example, the request may be to restore a logical device to its previous snapshot and the restrictions may indicate to not perform the restore if the logical device is in an active mirroring relationship, a source of an active migration, or if the snapshot has expired. The request may be to destroy the logical device which is included in a consistency group and a restriction indicates that a logical device included in a consistency group may not be destroyed.

Abstract: Described are techniques for processing a request received at a business logic provider from a client. The request includes a snapshot identifier uniquely identifying a snapshot having a snapshot object in a first data model and requests properties of a storage device providing a data source for the snapshot identified by the snapshot identifier. The snapshot object is requested from, and returned by, a snapshot provider. The business logic provider requests, from a storage device provider using a first property of the snapshot object, a storage device object corresponding to the storage device identified by the first property. The storage device provider returns the storage device object to the business logic provider. The storage device object is then returned to the client.

Abstract: Social networks by their nature are networks of people and their user-generated content. This content is usually un-curated and does not follow a general theme. This makes it very hard for people to find information they seek in these networks. When it comes to business centric communities, they need a better network where content is curated, tagged, and indexed properly thus making it easy for users to find the relevant content. SharePract is a community of curated, tagged, and indexed companies along with curated, tagged, and indexed content, making it easy for companies to find their business partners, clients, vendors, service providers, legal documents, marketing reports, business videos, webinars, and much more.

Abstract: An information system has a database having distinct object structures that store information about corresponding operating components of the system. An association structure describes many-to-many relationships among the components. Targeted requests for information about first objects are satisfied by obtaining information from a first object structure as well as information about associated objects from the second object structure. Bulk requests for information about a large number of objects are handled differently, using a lookup structure that associates the first objects with limited information of the associated second objects, such as externally visible names which would be meaningful to a requestor. Thus response can efficiently provide full information about all first objects as well as limited but useful information about all associated second objects, avoiding traversals of the association structure and corresponding interrogations of the second object structure.

Abstract: A method for text recognition includes generating a number of text hypotheses for an image, for example, using an HMM based approach using fixed-width analysis features. For each text hypothesis, one or more segmentations are generated and scored at the segmental level, for example, according to character or character group segments of the text hypothesis. In some embodiments, multiple alternative segmentations are considered for each text hypothesis. In some examples, scores determined in generating the text hypothesis and the segmental score are combined to select an overall text recognition of the image.

Abstract: A method for text recognition includes generating a number of text hypotheses for an image, for example, using an HMM based approach using fixed-width analysis features. For each text hypothesis, one or more segmentations are generated and scored at the segmental level, for example, according to character or character group segments of the text hypothesis. In some embodiments, multiple alternative segmentations are considered for each text hypothesis. In some examples, scores determined in generating the text hypothesis and the segmental score are combined to select an overall text recognition of the image.