US 11,947,926 B2
Discourse-level text optimization based on artificial intelligence planning
Akihiro Kishimoto, Castleknock (IE); Beat Buesser, Ashtown (IE); Bei Chen, Blanchardstown (IE); and Yufang Hou, Dublin (IE)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Sep. 25, 2020, as Appl. No. 17/032,264.
Prior Publication US 2022/0100969 A1, Mar. 31, 2022
Int. Cl. G06F 40/58 (2020.01); G06F 40/263 (2020.01); G06N 5/01 (2023.01)
CPC G06F 40/58 (2020.01) [G06F 40/263 (2020.01); G06N 5/01 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving, by one or more processors, an initial text in a first language;
applying, by one or more processors, one or more operators to modify the initial text, wherein the initial text is modified based on a candidate text;
evaluating, by one or more processors, the modified text using a scoring function, wherein evaluating is based on applying one or more user assigned weights associated with one or more example modified texts;
generating one or more facts from a knowledge database using the one or more user assigned weights;
determining, by one or more processors, whether a score generated from the scoring function on the modified text is above a predefined threshold, wherein the predefined threshold includes a runtime of an iterative deepening depth-first search, a maximum number of planning operators applied to the initial text, and a score of a target text; and
in response to determining the score is above the predefined threshold, outputting, by one or more processors, the modified text.