Automated File Attachments Based on User Intent
The first patent describes a technology that will automatically attach files to documents based on their relevance and the user's intent. This tool is particularly suited for Microsoft's suite of mailing and file management apps, including Outlook, Microsoft Teams, and OneDrive. The technology predicts the user's intention to attach a file and, upon confirmation, offers a list of relevant files for potential attachment. This process uses an intent model, an embedding model, and an N-gram model, combined with limited seed queries, to determine the user's intent and rank files based on their relevance. The primary beneficiaries of this innovation are expected to be Outlook users, but the technology's utility could extend to other apps both within and outside Microsoft's ecosystem.
Innovative Auto File-Editing Feature
The second patent focuses on a unique auto file-editing feature. This technology uses user input to effortlessly edit files, reminiscent of the capabilities of Windows Copilot and Microsoft 365 Copilot. The system learns from past file-editing experiences to suggest necessary edits when a new file interacts with a device. It operates by generating current context information, which includes an input message indicating the user's editing intention and the selected file content that needs modification. A pattern-completion engine, powered by a machine-trained model based on revision history, then predicts the necessary changes. This ensures that the edits are accurate, meet the user's objectives, and adhere to performance metrics.
The potential applications of this auto file-editing technology are vast. While it could be a feature in Copilot, there's also the possibility of integrating it into all Microsoft 365 applications, including Windows and Microsoft Teams. It could also work with readable storage data like external hard drives and SD cards.