Exploring the Thai Job Market Through the Lens of Natural Language Processing and Machine Learning
Abstract
In recent decades, the Beveridge curve, which demonstrates a relationship between unemployment and vacancies, has emerged as a central organizing framework for understanding of labour markets – both for academic as well as central banks. The absence of consistent of the data in Thailand is a fundamental drawback in the utilisation of this important indicator. Data from online job platforms presents an alternative opportunity. However, the first and necessary step is to develop a process that can structure and standardise such data. In this paper, we develop an algorithm that standardise the high-frequency data from job websites, which consists of manually written job titles from major online job posting websites in Thailand (in Thai and English languages) into the International Standard Classification of Occupations codes (ISCO-2008), up to 4-digit level. With Natural Language Processing and machine learning techniques, our methodology automates the process to efficiently deal with the volume and velocity nature of the data. Our approach not only carves a new path for comprehending labour market trends, but also enhances the capacity for monitoring labour market behaviours with higher precision and timeliness. Most of all, it offers a pivotal shift towards leveraging real-time, rich online job postings.