Home // DBKDA 2021, The Thirteenth International Conference on Advances in Databases, Knowledge, and Data Applications // View article

Analysing CV Corpus for Finding Suitable Candidates using Knowledge Graph and BERT

Authors:
Yan Wang
Yacine Allouache
Christian Joubert

Keywords: CV; CMAP; Knowledge Graph; BERT; TF-IDF Vectorizer; NER; MLM; HR-analytics; Job Matching.

Abstract:
Recruiter and candidate are the two main roles in the process of employment. Even though there is an abundance of job openings and a scarcity of qualified candidates to fill those openings, the objective is to offer only the profiles that fit the requirements of clients. Bidirectional Encoder Representations from Transformers (BERT) have been proposed in 2018 to better understand client searches. The challenges today are the frequent evolution of the experiences in Curriculum Vitae (CV) and the need of adaptive data for the specific staffing tasks of BERT. In this paper, we present an approach of ranking candidates based on competence keywords. There are four stages. First, we use Term Frequency–Inverse Document Frequency (TF-IDF) Vectorizer to calculate the score of matching between a competence keyword and a corpus of CVs. Second, we apply the Weighted Average Method to calculate a global score of CV based on two types of competence keywords – function and specialty. Third, we construct a Knowledge Graph (KG) from the structured Competence Map (CMAP), which can classify the relationships of bidirectional association and aggregation. At last, we propose to use the Named-Entity Recognition (NER) and Masked Language Modeling (MLM) of BERT to better identify tokens from the input inquiries of the client. The experiments are using the CVs from the HR (Human Resource) management system of Altran.

Pages: 26 to 31

Copyright: Copyright (c) IARIA, 2021

Publication date: May 30, 2021

Published in: conference

ISSN: 2308-4332

ISBN: 978-1-61208-857-0

Location: Valencia, Spain

Dates: from May 30, 2021 to June 3, 2021