US 12,248,451 B2
Machine-learning based automated document integration into genealogical trees
Jack Reese, Lindon, UT (US); Luca Lugini, San Mateo, CA (US); Yingrui Yang, San Mateo, CA (US); Simon Chu, Mountain View, CA (US); and Gopalkrishna Balkrishna Veni, Lehi, UT (US)
Assigned to Ancestry.com Operations Inc., Lehi, UT (US)
Filed by Ancestry.com Operations Inc., Lehi, UT (US)
Filed on Mar. 15, 2023, as Appl. No. 18/121,997.
Claims priority of provisional application 63/319,847, filed on Mar. 15, 2022.
Prior Publication US 2023/0325373 A1, Oct. 12, 2023
Int. Cl. G06F 16/22 (2019.01); G06F 40/295 (2020.01); G06V 30/32 (2022.01); G06V 30/413 (2022.01); G06V 30/416 (2022.01)
CPC G06F 16/2246 (2019.01) [G06F 40/295 (2020.01); G06V 30/333 (2022.01); G06V 30/413 (2022.01); G06V 30/416 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
receiving a genealogical record, the genealogical record including one or more images of a historical physical record;
segmenting the genealogical record into one or more segments, wherein segmenting the genealogical record comprises:
accessing a machine learning model trained over a plurality of training samples, each sample comprising at least a segment of document and a label indicating a targeted type of document structure associated with the segment of the document, wherein the machine learning model is trained by:
predicting a type of document structure for each of the plurality of training samples; and
applying an objective function to reduce a difference between the predicted type and the labeled targeted type of document structure of the respective training sample;
applying the machine learning model to the one or more images of the genealogical record to identify one or more types of document structures in the one or more images of the genealogical record; and
segmenting the one or more images based in part on one of the identified types of document structures;
extracting a plurality of named entities in one of the segments;
detecting a relationship among the plurality of named entities extracted in one of the segments;
adding one or more named entities in the plurality of named entities to a genealogical tree; and
causing to display the genealogical tree with added named entities at a graphical user interface, the genealogical tree including an access to at least one of the images of the historical physical record.