CONTEXTUALLY-AWARE RECOMMENDATIONS FOR ASSISTING USERS WITH TASK COMPLETION

摘要

A system for assisting users in making progress towards the completion of a task by recommending actions is provided. The techniques disclosed herein enable a system to receive user input signals, such as a voice input, a text input or any other type of input comprising phrases indicating a task. The system can then automatically generate recommendations for directing the user to complete the task. The techniques disclosed herein can also identify a contextually-relevant application that is most suitable for completing the identified task. For instance, a user may have several messaging applications. Based on a context of the user's input, and other historical data utilized by one or more machine learning engines, a system can recommend an appropriate application, or provide a ranking of the applications, suitable for completing an identified task. The techniques disclosed herein also employ machine learning engines to improve accuracy of the recommendations over time.