US 11,675,582 B2
Neural networks to identify source code
Balasubramanian Manivasagam, Bengaluru (IN); Thomas Domin, Bangalore (IN); Sakthimurugan Arumugam, Bangalore (IN); Thangadurai Muthusamy, Bangalore (IN); and Raja Sreenivasan, Bangalore (IN)
Assigned to KYNDRYL, INC., New York, NY (US)
Filed by Kyndryl, Inc., New York, NY (US)
Filed on Jul. 15, 2021, as Appl. No. 17/376,610.
Prior Publication US 2023/0016897 A1, Jan. 19, 2023
Int. Cl. G06F 8/71 (2018.01); G06F 8/10 (2018.01); G06F 40/20 (2020.01); G06N 20/00 (2019.01); G06F 18/2415 (2023.01)
CPC G06F 8/71 (2013.01) [G06F 8/10 (2013.01); G06F 18/24155 (2023.01); G06F 40/20 (2020.01); G06N 20/00 (2019.01)] 15 Claims
OG exemplary drawing
 
1. A computer-implemented method for searching source code using definitions for requirements comprising:
extracting search elements from requirement definitions of a requirement management tool for managing a project;
matching the search elements to identify source code from source code repositories, wherein machine learning correlates the requirement definitions to source code subject matter, the matching of search elements to identify source code comprises an artificial intelligence based solution fetcher that uses a naive Bayes classifier to classify code as completely matched or semi matches to the requirement definitions in combination with a Kernel density estimation to increase the accuracy of the naive Bayes classifier; and
confirming the source code that is matching the search elements meets the requirement definitions of the requirement management tool.