Handy Geraldy, Rizka Amalia Farentina, Fransisca Angelina Dirk

Identifying Digital Skills from Job Vacancy Using Text Classification and Named Entity Recognition: Case Study on Jobstreet Portal

Introduction

Identifying digital skills from job vacancy using text classification and named entity recognition: case study on jobstreet portal. Identify digital skills in Jobstreet job vacancies using text classification (XGBoost F1-score 94.33%) & NER. Maps digital talent demand & reveals top skills.

Abstract

Technological advancements have significantly reshaped the nature of work. A survey conducted by APINDO indicates that technology adoption contributed to elevated layoff rates during January-March 2025. Meanwhile, McKinsey & Company states that Indonesia will need 9 million digital talents (2014-2030). This study maps digital talent demand by classifying job vacancies data from Jobstreet based on digital skill levels (digital, semi-digital, and non-digital) and identifying the most frequently mentioned digital skills. The XGBoost achieves the best performance with an F1-score of 94.33%, outperforming SVM, logistic regression, and random forest. The study has provided an overview of job vacancy classifications based on the level of digital skills required. The XGBoost results indicate that 53,1% of job vacancies are classified as non-digital jobs. Furthermore, the NER model successfully identified skill entities in digital job vacancies, revealed that “communication”, “problem solving”, “software”, “design”, “SQL”, and “programming” were the most frequently mentioned skills.

Review

The submitted paper, "Identifying Digital Skills from Job Vacancy Using Text Classification and Named Entity Recognition: Case Study on Jobstreet Portal," addresses a highly pertinent and timely topic: the evolving demand for digital skills in the labor market. Given the rapid technological advancements and their documented impact on employment dynamics, as highlighted by references to APINDO and McKinsey & Company, understanding the current landscape of digital talent demand is crucial for policymakers, educators, and job seekers alike. This study leverages computational linguistics techniques to systematically analyze job vacancies, offering valuable insights into skill requirements and the prevalence of digital roles within a significant online job portal. The methodology employed by the authors is robust and well-suited for the stated objectives. The study utilizes text classification to categorize job vacancies into digital, semi-digital, and non-digital skill levels, and subsequently applies Named Entity Recognition (NER) to pinpoint specific digital skills. The reported performance of the XGBoost model, achieving an impressive F1-score of 94.33% and outperforming other established machine learning algorithms like SVM and random forest, attests to the effectiveness of their classification approach. A key finding is the significant proportion (53.1%) of jobs classified as non-digital, alongside the successful identification of frequently mentioned skills such as "communication," "problem solving," "software," "design," "SQL," and "programming," which collectively paint a comprehensive picture of current job market demands in the specified context of the Jobstreet portal. While the study presents compelling results, a more detailed elaboration on the criteria used to define "digital," "semi-digital," and "non-digital" job categories would further enhance its methodological transparency and allow for better interpretation of the 53.1% "non-digital" finding. Additionally, the inclusion of general skills like "communication" and "problem solving" among frequently mentioned "digital skills" could be further discussed, clarifying their intersection with or distinction from purely technical competencies. Future work could potentially explore the longitudinal trends of these skill demands, or validate the findings across different job portals or regional contexts to assess generalizability beyond this specific case study. Overall, this paper makes a significant contribution to understanding the dynamics of digital skill requirements in the contemporary job market, providing a solid foundation for subsequent research and practical applications.

Full Text

Comments

You need to be logged in to post a comment.