Predicting potentially relevant topics based on searched / created digital media files

摘要

Embodiments for utilizing digital media files searched and / or created by a user to anticipate / determine topics that are potentially relevant to the user are described herein. In various embodiments, digital media files created and / or retrieved by the user using the client device are applied as input across a trained machine learning model that is local to the client device in some cases and are digital. It can produce output showing the objects found in the media file. Data representing the indicated object may be provided to the remote computing system without providing the digital media file itself. In some embodiments, the information associated with the indicated object may be retrieved and proactively output to the user. In some embodiments, the frequency with which objects appear across the corpus of digital media files can be considered when determining the likelihood that the detected objects are potentially relevant to the user.