Microsoft patents system that uses AI to organize media files into a connected database
This filing describes a method for processing batches of media files by first removing duplicates and ranking the remaining items. The system uses a Large Language Model to generate descriptions of the content and creates a graph database that links media files to specific entities and metadata.
In the filing’s own words
In various examples there is a computer-implemented method, comprising ingesting into a memory a batch of media items. Duplicate media items in the batch are identified. For each duplicate media item, the duplicate media item is ranked with respect to duplicates of the media item. Filtering the duplicate media items is performed using the ranks of the duplicate media items to obtain filtered media items. Metadata of individual ones of the filtered media items is computed. Individual ones of the filtered media items are labelled with part of the metadata. For each filtered media item having metadata and a label, input into a Large Language Model, LLM, the media item, the label, the metadata and a request, and receive from the LLM a description of content depicted in the media item. A graph database of the filtered media items is created using the descriptions, the metadata, and the labels.
What exactly they claim (claim 1)
1. A computer-implemented method, comprising: ingesting into a memory a batch of media items; identifying, for a media item of the batch of media items, duplicate media items in the batch; for each duplicate media item, ranking the duplicate media item with respect to duplicates of the media item; filtering the duplicate media items based on the ranks of the duplicate media items to obtain filtered media items; computing metadata of individual ones of the filtered media items; labelling individual ones of the filtered media items with part of the metadata; for each filtered media item having metadata and a label: inputting, into a Large Language Model (LLM), the media item, the label, the metadata and a request; and receiving from the LLM a description of content depicted in the media item; and creating a graph database of the filtered media items using the descriptions, the metadata, and the labels, wherein creating the graph database comprises: for each filtered media item of the filtered media items: creating a first node representing the filtered media item; extracting, from the filtered media item, one or more entities depicted in the filtered media item; creating a second node representing the one or more entities extracted from the filtered media item; annotating the second node with metadata of the filtered media item; and connecting the first node and the annotated second node in the graph database.
Facts
- Company
- Microsoft (Microsoft Technology Licensing, LLC)
- Official title
- MEDIA ITEM GRAPH DATABASE
- Publication number
- US20260288866A1
- Application number
- 19085335
- Filed
- Mar 20, 2025
- Published
- Sep 24, 2026
- Inventors
- 8
- Claims
- 20
Classified under
G06F 16/9024 Physics & computingG06F 16/435 Physics & computingG06F 16/438 Physics & computingG06F 21/577 Physics & computingMore from Microsoft in AI & software
- Microsoft patents method to check AI answers against verified facts Oct 20, 2026
- Microsoft patents AI that presents PowerPoint slides for you Sep 3, 2026
- Microsoft patents method to check if LLM answers are based on real data Sep 10, 2026
- Microsoft patents system to blend multiple AI agents into one chat Oct 1, 2026
- Microsoft patents system that answers questions using your screen history Oct 15, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.