US 11,983,497 B2
Identification and personalized protection of text data using shapley values
Matthew John Schneider, Philadelphia, PA (US); and Shawn Mankad, Ithaca, NY (US)
Assigned to Drexel University, Philadelphia, PA (US)
Filed by Drexel University, Philadelphia, PA (US); and Cornell University, Ithaca, NY (US)
Filed on Nov. 20, 2020, as Appl. No. 16/953,516.
Claims priority of provisional application 62/937,851, filed on Nov. 20, 2019.
Prior Publication US 2021/0165965 A1, Jun. 3, 2021
Int. Cl. G06F 40/00 (2020.01); G06F 40/295 (2020.01); G06N 7/01 (2023.01); G06V 30/414 (2022.01); G06V 30/416 (2022.01)
CPC G06F 40/295 (2020.01) [G06N 7/01 (2023.01); G06V 30/414 (2022.01); G06V 30/416 (2022.01)] 7 Claims
OG exemplary drawing
 
1. A method for identifying the author of a written work, comprising:
calculating a Shapley value using the formula

OG Complex Work Unit Math
where for document i, variable j, and author k, where S is a subset of features used in the model, m the number of features, pk(S) is the predicted probability for author k using feature values in S;
approximating a % value of a true author by solving for Equation (1) using a Monte-Carlo sampling approximation method.