US 11,886,470 B2
Apparatus and method for aggregating and evaluating multimodal, time-varying entities
Adam Oliner, San Francisco, CA (US); Maria Kazandjieva, Menlo Park, CA (US); Eric Schkufza, Oakland, CA (US); Mher Hakobyan, Mountain View, CA (US); Irina Calciu, Palo Alto, CA (US); Brian Calvert, San Francisco, CA (US); and Daniel Woolridge, Los Angeles, CA (US)
Assigned to Graft, Inc., San Francisco, CA (US)
Filed by Graft, Inc., San Francisco, CA (US)
Filed on Feb. 23, 2022, as Appl. No. 17/678,942.
Application 17/678,942 is a continuation in part of application No. 17/488,043, filed on Sep. 28, 2021.
Claims priority of provisional application 63/216,431, filed on Jun. 29, 2021.
Prior Publication US 2023/0069958 A1, Mar. 9, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/31 (2019.01); G06F 16/33 (2019.01)
CPC G06F 16/31 (2019.01) [G06F 16/3347 (2019.01)] 11 Claims
OG exemplary drawing
 
1. A non-transitory computer readable storage medium with instructions executed by a processor to:
receive from a network connection different sources of unstructured data, wherein the unstructured data has multiple modes of semantically distinct data types and the unstructured data has time-varying data instances aggregated over time;
form an entity combining different sources of the unstructured data;
create a representation for the entity, wherein the representation includes embeddings that are numeric vectors computed using machine learning embedding models;
infer a property of the entity to form an enrichment with an attribute characterizing the entity;
repeat the operations to receive, form, create and infer to form an aggregation of multimodal, time-varying entities and a corresponding index of individual entities and corresponding embeddings; and
perform proximity searches on embeddings within the index.