US 11,995,612 B2
Knowledge engine using machine learning and predictive modeling for optimizing recruitment management systems
Mahe Bayireddi, Warrington, PA (US); Hari Bayireddy, Warrington, PA (US); Sivanand Akella, Hyderabad (IN); Suresh Babu Devineni, Hyderabad (IN); S. S. S. Venkateswara Rao Alapati, Visakhapatnam (IN); Shivaramakrishna Nyshadham, Hyderabad (IN); and Ilya Goldin, Pittsburgh, PA (US)
Assigned to Phenom People, Ambler, PA (US)
Filed by Phenom People, Ambler, PA (US)
Filed on Apr. 8, 2020, as Appl. No. 16/843,805.
Claims priority of provisional application 62/830,680, filed on Apr. 8, 2019.
Prior Publication US 2020/0320483 A1, Oct. 8, 2020
Int. Cl. G06Q 10/1053 (2023.01); G06N 20/00 (2019.01)
CPC G06Q 10/1053 (2013.01) [G06N 20/00 (2019.01)] 16 Claims
OG exemplary drawing
 
1. A method for creating and updating a weighted knowledge graph, the method comprising:
receiving a request for a potential candidate for a job, wherein the request for the potential candidate is a passive interaction;
sending a response to the request for the potential candidate for the job based on the request for the potential candidate being the passive interaction, wherein the response comprises a request that an operator associated with the request for the potential candidate for the job authenticate an identity associated with the operator;
receiving the identity associated with the operator;
retrieving a weighted knowledge graph associated with the job, wherein the job is associated with an employer;
retrieving a weighted knowledge graph associated with the employer;
determining, by a knowledge engine, a plurality of candidates for the job based on a respective weighted knowledge graph associated with each of the plurality of candidates, the weighted knowledge graph associated with the employer, and the knowledge graph associated with the job, wherein the knowledge engine comprises a plurality of clusters, and wherein each of the plurality of clusters is associated with one or more gradients;
receiving a plurality of data files, each of the plurality of data files comprising knowledge entities relating to each of the plurality of candidates;
generating vectors associated with each of the plurality of data files;
associating each of the knowledge entities with at least one of a plurality of categories by passing the vectors associated with each of the plurality of data files to the knowledge engine, wherein the plurality of categories comprises a technical skills category, a job responsibilities category, a soft-skills category, and an educational qualification category, and wherein each of the plurality of categories is associated with a cluster of the plurality of clusters;
updating, using the knowledge engine, the weighted knowledge graph associated with the job and the weighted knowledge graphs associated with each of the plurality of candidates based on each of the knowledge entities associated with the at least one of the plurality of categories, wherein each of the knowledge entities are further associated with respective weighted relationships, each of the respective weighted relationships being associated with a relationship strength, a relationship type, and a relationship direction, wherein the relationship strength, the relationship type and the relationship direction indicate a likelihood of a personal connection between each of the plurality of candidates and the job;
determining, by the knowledge engine, a weighted relationship between each of the plurality of candidates and the job based on each of the respective weighted relationships;
updating the knowledge engine based on the determined weighted relationship between each of the plurality of candidates and the job, wherein updating the knowledge engine comprises updating the one or more gradients associated with each of the plurality of clusters using backpropagation;
generating, based on the updated knowledge engine, a list of ranked candidates for the job based on the updated weighted knowledge graph associated with the job and the updated weighted knowledge graphs associated with each of the plurality of candidates; and
sending a second response to the request for the potential candidate for the job, wherein the second response displays the generated list of ranked candidates and an indication of a likelihood of a personal connection between each of the ranked candidates and the job.