US 12,190,059 B2
Systems and methods for deviation detection, information extraction and obligation deviation detection
Sally Gao, Boston, MA (US); Hella-Franziska Hoffmann, London (GB); Nina Hristozova, Zurich (CH); Elizabeth Roman, Somerville, MA (US); Nicolai Pogrebnyakov, Ontario (CA); Yue Feng, Ontario (CA); Masoud Makrehchi, Ontario (CA); Tate Sterling Avery, Ontario (CA); Shohreh Shaghaghian, Ontario (CA); and Borna Jafarpour, Ontario (CA)
Assigned to Thomson Reuters Enterprise Centre GmbH, Zug (CH)
Filed by Thomson Reuters Enterprise Centre GmbH, Zug (CH)
Filed on Jan. 23, 2021, as Appl. No. 17/156,570.
Claims priority of provisional application 62/975,514, filed on Feb. 12, 2020.
Claims priority of provisional application 62/965,516, filed on Jan. 24, 2020.
Claims priority of provisional application 62/965,523, filed on Jan. 24, 2020.
Claims priority of provisional application 62/965,520, filed on Jan. 24, 2020.
Prior Publication US 2021/0350080 A1, Nov. 11, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 40/289 (2020.01); G06F 3/0481 (2022.01); G06F 40/109 (2020.01); G06F 40/137 (2020.01); G06F 40/166 (2020.01); G06F 40/232 (2020.01); G06F 40/242 (2020.01); G06F 40/258 (2020.01); G06F 40/279 (2020.01); G06F 40/284 (2020.01); G06V 30/416 (2022.01)
CPC G06F 40/279 (2020.01) [G06F 3/0481 (2013.01); G06F 40/109 (2020.01); G06F 40/137 (2020.01); G06F 40/166 (2020.01); G06F 40/232 (2020.01); G06F 40/242 (2020.01); G06F 40/258 (2020.01); G06F 40/284 (2020.01); G06F 40/289 (2020.01); G06V 30/416 (2022.01)] 12 Claims
OG exemplary drawing
 
1. A method for detecting deviations, comprising:
receiving a standard document and a review document;
splitting apart sentences in each of the standard document and review document;
converting the sentences in each of the standard and review document into respective feature sets for a classifier model;
providing the respective feature sets for the sentences in each of the standard and review document as input to a classifier model configured to provide a predicted similarity between a review sentence from the review document and a standard sentence from the standard document based on a pairwise sentence alignment between the review document and the standard document;
in response to the predicted similarity between a particular review sentence from the review document and a particular standard sentence from the standard document being above a defined similarity threshold, detecting deviations in the particular review sentence from the particular standard sentence; and
in response to detecting the deviations, causing display of visual data associated with the deviations via a user interface of a user computing device.