Abstract: Techniques and systems are described to improve computing device telemetry collection by automatic analysis and dynamic adjustment of telemetry collection rules. Certain implementations use a telemetry service to analyze telemetry messages and direct the adjusted rules at the client computing device, automatically and gradually refining telemetry collection as more relevant telemetry is received. Dynamic adjustments that can be made to the collection rules include, for example, changes to the collection frequency of telemetry messages, changes to the level of detail transmitted in the telemetry messages, and changes to the targets (e.g., applications, components, devices) from which to collect.

Abstract: A predictive contextual toolbar is presented that provides an identifiable region on which predicted commands can be surfaced. This user interface can be presented in cases where an indication of a request for a command (such as a formatting tool) is received, for example, while a user is interacting with a canvas of a productivity application. When conditions for surfacing a predictive command are met, the system can display an identifiable region on which predicted commands are surfaced according to context associated with the conditions for surfacing the predictive command. The identifiable region may be an extension of a mini toolbar. One or more dynamic predicted commands can be surfaced on the identifiable region to provide easily accessible, user specific, context related commands.

Abstract: Techniques and systems are described to improve computing device telemetry collection by automatic analysis and dynamic adjustment of telemetry collection rules. Certain implementations use a telemetry service to analyze telemetry messages and direct the adjusted rules at the client computing device, automatically and gradually refining telemetry collection as more relevant telemetry is received. Dynamic adjustments that can be made to the collection rules include, for example, changes to the collection frequency of telemetry messages, changes to the level of detail transmitted in the telemetry messages, and changes to the targets (e.g., applications, components, devices) from which to collect.

Abstract: Systems and techniques for facilitating and backing the surfacing of predicted commands within a user interface are disclosed. Commands to surface for an active user in productivity applications can be predicted using a personalized community model. The personalized community model is generated using a record of past actions the active user has taken along with the past actions of many users of the productivity application. The actions of the active user within the productivity application are monitored and used to select commands to surface.

Abstract: A predictive contextual toolbar is presented that provides an identifiable region on which predicted commands can be surfaced. This user interface can be presented in cases where an indication of a request for a command (such as a formatting tool) is received, for example, while a user is interacting with a canvas of a productivity application. When conditions for surfacing a predictive command are met, the system can display an identifiable region on which predicted commands are surfaced according to context associated with the conditions for surfacing the predictive command. The identifiable region may be an extension of a mini toolbar. One or more dynamic predicted commands can be surfaced on the identifiable region to provide easily accessible, user specific, context related commands.

Abstract: Optimization architecture for applications that uses a localization model in which localizable resources are separated from code. The resources are typically located in separate DLL files as binary blobs in a unique format. The format is designed for optimized performance and to accommodate various requirements of complex office applications. The binary blobs are produced by a resource compiler based on the format. The architecture includes a resource binary format that facilitates resource identification and lookup, optimization of memory paging by grouping and ordering resources according to runtime use, optimization by compressing resources with an optimal algorithm that is fast in decompression, fast reverse lookup of a resource identifier by resource content, resource grouping, resource substitution, branding, and resource runtime metadata (also referred to as user data).