US 11,934,445 B2
Automatic memory content item provisioning
Maheen Sohail, San Francisco, CA (US); Hyunbin Park, Palo Alto, CA (US); Xinyue Zhang, Mountain View, CA (US); and Tali Zvi, San Carlos, CA (US)
Assigned to Meta Platforms Technologies, LLC, Menlo Park, CA (US)
Filed by Meta Platforms Technologies, LLC, Menlo Park, CA (US)
Filed on Feb. 11, 2021, as Appl. No. 17/174,293.
Claims priority of provisional application 63/131,274, filed on Dec. 28, 2020.
Prior Publication US 2022/0207073 A1, Jun. 30, 2022
Int. Cl. G10L 15/00 (2013.01); G06F 16/438 (2019.01); G06N 3/08 (2023.01); G06Q 50/00 (2012.01); G10L 15/22 (2006.01); G10L 15/26 (2006.01)
CPC G06F 16/4393 (2019.01) [G06N 3/08 (2013.01); G06Q 50/01 (2013.01); G10L 15/22 (2013.01); G10L 15/26 (2013.01); G10L 2015/223 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for automatic presentation of memory content items, the method comprising:
obtaining natural language (NL) input;
computing match scores, wherein each match score indicates a match level directly between the NL input and one of multiple memory content items, and wherein each match score is based on a comparison directly between the NL input and one of the multiple memory content items, the comparison comprising:
segmenting the NL input into sections;
providing a particular memory content item and the sections to a machine learning model;
generating multiple sub-match scores, each computed by applying the machine learning model to the particular memory content item and to one of the sections; and
combining the sub-match scores, for the particular memory content item and the sections, into one of the match scores;
selecting one or more memory content items corresponding to one or more of the highest computed match scores; and
providing the selected one or more memory content items.