Abstract: Techniques for conversation recovery in a dialog management system are described. A system may determine, using dialog models, that a predicted action to be performed by a skill component is likely to result in an undesired response or that the skill component is unable to respond to a user input of a dialog session. Rather than informing the user that the skill component is unable to respond, the system may send data to the skill component to enable the skill component to determine a correct action responsive to the user input. The data may include an indication of the predicted action and/or entity data corresponding to the user input. The system may receive, from the skill component, response data corresponding to the user input, and may use the response data to update a dialog context for the dialog session and an inference engine of the dialog management system.

Abstract: User interactions with presented content can be tracked and a corresponding user behavior identified. An enhancement trigger can be detected by comparing the user behavior with a pattern of behavior associated with the enhancement trigger. A modified experience can be provided at the user device. The modified experience can include a user interface for interacting with a chat bot.

Abstract: Disclosed are various embodiments for recommending a list of digital media to a user based upon digital media that the user is currently consuming. The list of recommendations is generated from an analysis of the digital media. Subsequently, the list of recommendations is stored on the user's device, and the user is alerted to the list's existence through the rendering of a notification on the user's display. Upon selection of an item from the list of recommendations, the selected digital media is rendered on the display of the user's device within the context of the user's application.

Abstract: Widgets or other code modules for applications written in structured languages such as HTML can be tested advantageously by separating the extraction and validation logic. To test the functionality of a widget in response to an action, widget features of interest are identified. Context and feature nodes for these features are extracted from a page tree structure to generate a property vector that is representative of the state of the widget. After performing the action, context nodes for the features are attached to the new page tree for the newly loaded page, and an observed property vector is generated that represents the state of the widget after the action. The previous property vector is used with updated state information to generate an expected property vector. The observed property vector is compared with the expected property vector, and if the vectors match the validation is claimed to be successful.

Abstract: Widgets or other code modules for applications written in structured languages such as HTML can be tested advantageously by separating the extraction and validation logic. To test the functionality of a widget in response to an action, widget features of interest are identified. Context and feature nodes for these features are extracted from a page tree structure to generate a property vector that is representative of the state of the widget. After performing the action, context nodes for the features are attached to the new page tree for the newly loaded page, and an observed property vector is generated that represents the state of the widget after the action. The previous property vector is used with updated state information to generate an expected property vector. The observed property vector is compared with the expected property vector, and if the vectors match the validation is claimed to be successful.

Abstract: A peer to peer audio communication screen in system with adaptive bandwidth control. The peer-to-peer system can be used, for example, in classroom applications. More interactive peers are moved to a position where they will receive lower latency from the peer-to-peer system. This may be done by maintaining a score indicative of the amount of interactivity, and then using that score to move a peer either close to the source of audio packets, e.g., the core or further from that source.