US 11,874,868 B2
Generating and presenting multi-dimensional representations for complex entities
Robin Abraham, Redmond, WA (US); Leo Betthauser, Kirkland, WA (US); Ziyao Li, Kirkland, WA (US); Jing Tian, Redmond, WA (US); Xiaofei Zeng, Redmond, WA (US); Maurice Diesendruck, Bellevue, WA (US); Andy Daniel Martinez, Redmond, WA (US); Min Xiao, Bothell, WA (US); Liang Du, Redmond, WA (US); Pramod Kumar Sharma, Seattle, WA (US); and Natalia Larios Delgado, Kirkland, WA (US)
Assigned to Microsoft Tech LLCnology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jun. 14, 2021, as Appl. No. 17/347,097.
Prior Publication US 2022/0398274 A1, Dec. 15, 2022
Int. Cl. G06F 16/00 (2019.01); G06F 16/35 (2019.01); G06F 16/45 (2019.01); G06F 16/483 (2019.01); G06F 16/44 (2019.01); G06F 16/383 (2019.01)
CPC G06F 16/358 (2019.01) [G06F 16/383 (2019.01); G06F 16/44 (2019.01); G06F 16/45 (2019.01); G06F 16/483 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
identifying a collection of digital content items including content associated with a plurality of instances of one or more entities;
causing a plurality of attribute models to be applied to the collection of digital content items to generate attribute signals associated with the collection of digital content items, wherein the attribute signals include a combination of one or more atomic attribute signals and one or more deep learned attribute signals;
generating an embedding index including a collection of embeddings for the plurality of instances of the one or more entities, the collection of embeddings including multi-dimensional representations of the plurality of instances based on the combination of one or more atomic attribute signals and one or more deep learned attribute signals generated by the plurality of attribute models, the multi-dimensional representations of the plurality of instances including multi-dimensional vectors having numeric values representative of the attribute signals, wherein generating the embedding index includes determining placements of the multi-dimensional vectors within a multi-dimensional space; and
providing an interactive presentation of the embedding index on a graphical user interface of a client device, the presentation including an indication of one or more embedding clusters that are traversable via the interactive presentation by a user of the client device, the one or more embedding clusters including groupings of similar instances from the plurality of instances based on proximity of the determined placement of the multi-dimensional representations within the multi-dimensional space of the embedding index.