Two Stage Job Title Identification System for Online Job Advertisements

The exponential growth of online job advertisements has created both opportunities and challenges in automated job market analysis.A key component of such analysis is the accurate identification of job titles from often noisy, unstructured, and inconsistent advertisement texts.This paper presents a novel Two-Stage Job Title Identification System specifically designed for online job advertisements.In the first stage, the system employs rule-based and machine learning methods to extract candidate job title phrases from the advertisement body.This includes filtering and cleaning operations to handle variations, abbreviations, and irrelevant terms.The second stage involves a deep learning-based classifier that evaluates the extracted phrases and determines the most probable job title by leveraging contextual embeddings.This two-tiered approach combines the precision of pattern-based extraction with the semantic richness of modern NLP models.The system is trained and tested on a large-scale dataset of job postings collected from major employment platforms, showing significant improvements in accuracy and robustness over traditional single-stage models.The modular design allows for adaptability across industries and regions, offering a practical solution for HR analytics, labor market research, and job recommendation systems.The proposed methodology not only improves identification accuracy but also facilitates better alignment between job seekers and employers in the digital job ecosystem.